From 36d22d82aa202bb199967e9512281e9a53db42c9 Mon Sep 17 00:00:00 2001 From: Daniel Baumann Date: Sun, 7 Apr 2024 21:33:14 +0200 Subject: Adding upstream version 115.7.0esr. Signed-off-by: Daniel Baumann --- .../modules/audio_coding/codecs/opus/DEPS | 5 + .../codecs/opus/audio_coder_opus_common.cc | 52 ++ .../codecs/opus/audio_coder_opus_common.h | 89 ++ .../opus/audio_decoder_multi_channel_opus_impl.cc | 182 ++++ .../opus/audio_decoder_multi_channel_opus_impl.h | 74 ++ .../audio_decoder_multi_channel_opus_unittest.cc | 148 ++++ .../audio_coding/codecs/opus/audio_decoder_opus.cc | 128 +++ .../audio_coding/codecs/opus/audio_decoder_opus.h | 64 ++ .../opus/audio_encoder_multi_channel_opus_impl.cc | 366 ++++++++ .../opus/audio_encoder_multi_channel_opus_impl.h | 92 ++ .../audio_encoder_multi_channel_opus_unittest.cc | 156 ++++ .../audio_coding/codecs/opus/audio_encoder_opus.cc | 824 +++++++++++++++++ .../audio_coding/codecs/opus/audio_encoder_opus.h | 184 ++++ .../codecs/opus/audio_encoder_opus_unittest.cc | 914 +++++++++++++++++++ .../codecs/opus/opus_bandwidth_unittest.cc | 152 ++++ .../codecs/opus/opus_complexity_unittest.cc | 105 +++ .../audio_coding/codecs/opus/opus_fec_test.cc | 248 ++++++ .../modules/audio_coding/codecs/opus/opus_inst.h | 43 + .../audio_coding/codecs/opus/opus_interface.cc | 881 ++++++++++++++++++ .../audio_coding/codecs/opus/opus_interface.h | 547 ++++++++++++ .../audio_coding/codecs/opus/opus_speed_test.cc | 147 ++++ .../audio_coding/codecs/opus/opus_unittest.cc | 979 +++++++++++++++++++++ .../modules/audio_coding/codecs/opus/test/BUILD.gn | 55 ++ .../codecs/opus/test/audio_ring_buffer.cc | 76 ++ .../codecs/opus/test/audio_ring_buffer.h | 57 ++ .../codecs/opus/test/audio_ring_buffer_unittest.cc | 111 +++ .../audio_coding/codecs/opus/test/blocker.cc | 215 +++++ .../audio_coding/codecs/opus/test/blocker.h | 127 +++ .../codecs/opus/test/blocker_unittest.cc | 293 ++++++ .../codecs/opus/test/lapped_transform.cc | 100 +++ .../codecs/opus/test/lapped_transform.h | 175 ++++ .../codecs/opus/test/lapped_transform_unittest.cc | 203 +++++ 32 files changed, 7792 insertions(+) create mode 100644 third_party/libwebrtc/modules/audio_coding/codecs/opus/DEPS create mode 100644 third_party/libwebrtc/modules/audio_coding/codecs/opus/audio_coder_opus_common.cc create mode 100644 third_party/libwebrtc/modules/audio_coding/codecs/opus/audio_coder_opus_common.h create mode 100644 third_party/libwebrtc/modules/audio_coding/codecs/opus/audio_decoder_multi_channel_opus_impl.cc create mode 100644 third_party/libwebrtc/modules/audio_coding/codecs/opus/audio_decoder_multi_channel_opus_impl.h create mode 100644 third_party/libwebrtc/modules/audio_coding/codecs/opus/audio_decoder_multi_channel_opus_unittest.cc create mode 100644 third_party/libwebrtc/modules/audio_coding/codecs/opus/audio_decoder_opus.cc create mode 100644 third_party/libwebrtc/modules/audio_coding/codecs/opus/audio_decoder_opus.h create mode 100644 third_party/libwebrtc/modules/audio_coding/codecs/opus/audio_encoder_multi_channel_opus_impl.cc create mode 100644 third_party/libwebrtc/modules/audio_coding/codecs/opus/audio_encoder_multi_channel_opus_impl.h create mode 100644 third_party/libwebrtc/modules/audio_coding/codecs/opus/audio_encoder_multi_channel_opus_unittest.cc create mode 100644 third_party/libwebrtc/modules/audio_coding/codecs/opus/audio_encoder_opus.cc create mode 100644 third_party/libwebrtc/modules/audio_coding/codecs/opus/audio_encoder_opus.h create mode 100644 third_party/libwebrtc/modules/audio_coding/codecs/opus/audio_encoder_opus_unittest.cc create mode 100644 third_party/libwebrtc/modules/audio_coding/codecs/opus/opus_bandwidth_unittest.cc create mode 100644 third_party/libwebrtc/modules/audio_coding/codecs/opus/opus_complexity_unittest.cc create mode 100644 third_party/libwebrtc/modules/audio_coding/codecs/opus/opus_fec_test.cc create mode 100644 third_party/libwebrtc/modules/audio_coding/codecs/opus/opus_inst.h create mode 100644 third_party/libwebrtc/modules/audio_coding/codecs/opus/opus_interface.cc create mode 100644 third_party/libwebrtc/modules/audio_coding/codecs/opus/opus_interface.h create mode 100644 third_party/libwebrtc/modules/audio_coding/codecs/opus/opus_speed_test.cc create mode 100644 third_party/libwebrtc/modules/audio_coding/codecs/opus/opus_unittest.cc create mode 100644 third_party/libwebrtc/modules/audio_coding/codecs/opus/test/BUILD.gn create mode 100644 third_party/libwebrtc/modules/audio_coding/codecs/opus/test/audio_ring_buffer.cc create mode 100644 third_party/libwebrtc/modules/audio_coding/codecs/opus/test/audio_ring_buffer.h create mode 100644 third_party/libwebrtc/modules/audio_coding/codecs/opus/test/audio_ring_buffer_unittest.cc create mode 100644 third_party/libwebrtc/modules/audio_coding/codecs/opus/test/blocker.cc create mode 100644 third_party/libwebrtc/modules/audio_coding/codecs/opus/test/blocker.h create mode 100644 third_party/libwebrtc/modules/audio_coding/codecs/opus/test/blocker_unittest.cc create mode 100644 third_party/libwebrtc/modules/audio_coding/codecs/opus/test/lapped_transform.cc create mode 100644 third_party/libwebrtc/modules/audio_coding/codecs/opus/test/lapped_transform.h create mode 100644 third_party/libwebrtc/modules/audio_coding/codecs/opus/test/lapped_transform_unittest.cc (limited to 'third_party/libwebrtc/modules/audio_coding/codecs/opus') diff --git a/third_party/libwebrtc/modules/audio_coding/codecs/opus/DEPS b/third_party/libwebrtc/modules/audio_coding/codecs/opus/DEPS new file mode 100644 index 0000000000..c2530726ad --- /dev/null +++ b/third_party/libwebrtc/modules/audio_coding/codecs/opus/DEPS @@ -0,0 +1,5 @@ +specific_include_rules = { + "opus_inst\.h": [ + "+third_party/opus", + ], +} diff --git a/third_party/libwebrtc/modules/audio_coding/codecs/opus/audio_coder_opus_common.cc b/third_party/libwebrtc/modules/audio_coding/codecs/opus/audio_coder_opus_common.cc new file mode 100644 index 0000000000..03c02186d0 --- /dev/null +++ b/third_party/libwebrtc/modules/audio_coding/codecs/opus/audio_coder_opus_common.cc @@ -0,0 +1,52 @@ +/* + * Copyright (c) 2019 The WebRTC project authors. All Rights Reserved. + * + * Use of this source code is governed by a BSD-style license + * that can be found in the LICENSE file in the root of the source + * tree. An additional intellectual property rights grant can be found + * in the file PATENTS. All contributing project authors may + * be found in the AUTHORS file in the root of the source tree. + */ + +#include "modules/audio_coding/codecs/opus/audio_coder_opus_common.h" + +#include "absl/strings/string_view.h" + +namespace webrtc { + +absl::optional GetFormatParameter(const SdpAudioFormat& format, + absl::string_view param) { + auto it = format.parameters.find(std::string(param)); + if (it == format.parameters.end()) + return absl::nullopt; + + return it->second; +} + +// Parses a comma-separated string "1,2,0,6" into a std::vector. +template <> +absl::optional> GetFormatParameter( + const SdpAudioFormat& format, + absl::string_view param) { + std::vector result; + const std::string comma_separated_list = + GetFormatParameter(format, param).value_or(""); + size_t pos = 0; + while (pos < comma_separated_list.size()) { + const size_t next_comma = comma_separated_list.find(',', pos); + const size_t distance_to_next_comma = next_comma == std::string::npos + ? std::string::npos + : (next_comma - pos); + auto substring_with_number = + comma_separated_list.substr(pos, distance_to_next_comma); + auto conv = rtc::StringToNumber(substring_with_number); + if (!conv.has_value()) { + return absl::nullopt; + } + result.push_back(*conv); + pos += substring_with_number.size() + 1; + } + return result; +} + +} // namespace webrtc diff --git a/third_party/libwebrtc/modules/audio_coding/codecs/opus/audio_coder_opus_common.h b/third_party/libwebrtc/modules/audio_coding/codecs/opus/audio_coder_opus_common.h new file mode 100644 index 0000000000..5ebb51b577 --- /dev/null +++ b/third_party/libwebrtc/modules/audio_coding/codecs/opus/audio_coder_opus_common.h @@ -0,0 +1,89 @@ +/* + * Copyright (c) 2019 The WebRTC project authors. All Rights Reserved. + * + * Use of this source code is governed by a BSD-style license + * that can be found in the LICENSE file in the root of the source + * tree. An additional intellectual property rights grant can be found + * in the file PATENTS. All contributing project authors may + * be found in the AUTHORS file in the root of the source tree. + */ + +#ifndef MODULES_AUDIO_CODING_CODECS_OPUS_AUDIO_CODER_OPUS_COMMON_H_ +#define MODULES_AUDIO_CODING_CODECS_OPUS_AUDIO_CODER_OPUS_COMMON_H_ + +#include +#include +#include + +#include "absl/strings/string_view.h" +#include "absl/types/optional.h" +#include "api/audio_codecs/audio_decoder.h" +#include "api/audio_codecs/audio_format.h" +#include "rtc_base/string_to_number.h" + +namespace webrtc { + +absl::optional GetFormatParameter(const SdpAudioFormat& format, + absl::string_view param); + +template +absl::optional GetFormatParameter(const SdpAudioFormat& format, + absl::string_view param) { + return rtc::StringToNumber(GetFormatParameter(format, param).value_or("")); +} + +template <> +absl::optional> GetFormatParameter( + const SdpAudioFormat& format, + absl::string_view param); + +class OpusFrame : public AudioDecoder::EncodedAudioFrame { + public: + OpusFrame(AudioDecoder* decoder, + rtc::Buffer&& payload, + bool is_primary_payload) + : decoder_(decoder), + payload_(std::move(payload)), + is_primary_payload_(is_primary_payload) {} + + size_t Duration() const override { + int ret; + if (is_primary_payload_) { + ret = decoder_->PacketDuration(payload_.data(), payload_.size()); + } else { + ret = decoder_->PacketDurationRedundant(payload_.data(), payload_.size()); + } + return (ret < 0) ? 0 : static_cast(ret); + } + + bool IsDtxPacket() const override { return payload_.size() <= 2; } + + absl::optional Decode( + rtc::ArrayView decoded) const override { + AudioDecoder::SpeechType speech_type = AudioDecoder::kSpeech; + int ret; + if (is_primary_payload_) { + ret = decoder_->Decode( + payload_.data(), payload_.size(), decoder_->SampleRateHz(), + decoded.size() * sizeof(int16_t), decoded.data(), &speech_type); + } else { + ret = decoder_->DecodeRedundant( + payload_.data(), payload_.size(), decoder_->SampleRateHz(), + decoded.size() * sizeof(int16_t), decoded.data(), &speech_type); + } + + if (ret < 0) + return absl::nullopt; + + return DecodeResult{static_cast(ret), speech_type}; + } + + private: + AudioDecoder* const decoder_; + const rtc::Buffer payload_; + const bool is_primary_payload_; +}; + +} // namespace webrtc + +#endif // MODULES_AUDIO_CODING_CODECS_OPUS_AUDIO_CODER_OPUS_COMMON_H_ diff --git a/third_party/libwebrtc/modules/audio_coding/codecs/opus/audio_decoder_multi_channel_opus_impl.cc b/third_party/libwebrtc/modules/audio_coding/codecs/opus/audio_decoder_multi_channel_opus_impl.cc new file mode 100644 index 0000000000..285ea89959 --- /dev/null +++ b/third_party/libwebrtc/modules/audio_coding/codecs/opus/audio_decoder_multi_channel_opus_impl.cc @@ -0,0 +1,182 @@ +/* + * Copyright (c) 2019 The WebRTC project authors. All Rights Reserved. + * + * Use of this source code is governed by a BSD-style license + * that can be found in the LICENSE file in the root of the source + * tree. An additional intellectual property rights grant can be found + * in the file PATENTS. All contributing project authors may + * be found in the AUTHORS file in the root of the source tree. + */ + +#include "modules/audio_coding/codecs/opus/audio_decoder_multi_channel_opus_impl.h" + +#include +#include +#include +#include +#include + +#include "absl/memory/memory.h" +#include "modules/audio_coding/codecs/opus/audio_coder_opus_common.h" +#include "rtc_base/string_to_number.h" + +namespace webrtc { + +std::unique_ptr +AudioDecoderMultiChannelOpusImpl::MakeAudioDecoder( + AudioDecoderMultiChannelOpusConfig config) { + if (!config.IsOk()) { + RTC_DCHECK_NOTREACHED(); + return nullptr; + } + // Fill the pointer with a working decoder through the C interface. This + // allocates memory. + OpusDecInst* dec_state = nullptr; + const int error = WebRtcOpus_MultistreamDecoderCreate( + &dec_state, config.num_channels, config.num_streams, + config.coupled_streams, config.channel_mapping.data()); + if (error != 0) { + return nullptr; + } + + // Pass the ownership to DecoderImpl. Not using 'make_unique' because the + // c-tor is private. + return std::unique_ptr( + new AudioDecoderMultiChannelOpusImpl(dec_state, config)); +} + +AudioDecoderMultiChannelOpusImpl::AudioDecoderMultiChannelOpusImpl( + OpusDecInst* dec_state, + AudioDecoderMultiChannelOpusConfig config) + : dec_state_(dec_state), config_(config) { + RTC_DCHECK(dec_state); + WebRtcOpus_DecoderInit(dec_state_); +} + +AudioDecoderMultiChannelOpusImpl::~AudioDecoderMultiChannelOpusImpl() { + WebRtcOpus_DecoderFree(dec_state_); +} + +absl::optional +AudioDecoderMultiChannelOpusImpl::SdpToConfig(const SdpAudioFormat& format) { + AudioDecoderMultiChannelOpusConfig config; + config.num_channels = format.num_channels; + auto num_streams = GetFormatParameter(format, "num_streams"); + if (!num_streams.has_value()) { + return absl::nullopt; + } + config.num_streams = *num_streams; + + auto coupled_streams = GetFormatParameter(format, "coupled_streams"); + if (!coupled_streams.has_value()) { + return absl::nullopt; + } + config.coupled_streams = *coupled_streams; + + auto channel_mapping = + GetFormatParameter>(format, "channel_mapping"); + if (!channel_mapping.has_value()) { + return absl::nullopt; + } + config.channel_mapping = *channel_mapping; + if (!config.IsOk()) { + return absl::nullopt; + } + return config; +} + +std::vector +AudioDecoderMultiChannelOpusImpl::ParsePayload(rtc::Buffer&& payload, + uint32_t timestamp) { + std::vector results; + + if (PacketHasFec(payload.data(), payload.size())) { + const int duration = + PacketDurationRedundant(payload.data(), payload.size()); + RTC_DCHECK_GE(duration, 0); + rtc::Buffer payload_copy(payload.data(), payload.size()); + std::unique_ptr fec_frame( + new OpusFrame(this, std::move(payload_copy), false)); + results.emplace_back(timestamp - duration, 1, std::move(fec_frame)); + } + std::unique_ptr frame( + new OpusFrame(this, std::move(payload), true)); + results.emplace_back(timestamp, 0, std::move(frame)); + return results; +} + +int AudioDecoderMultiChannelOpusImpl::DecodeInternal(const uint8_t* encoded, + size_t encoded_len, + int sample_rate_hz, + int16_t* decoded, + SpeechType* speech_type) { + RTC_DCHECK_EQ(sample_rate_hz, 48000); + int16_t temp_type = 1; // Default is speech. + int ret = + WebRtcOpus_Decode(dec_state_, encoded, encoded_len, decoded, &temp_type); + if (ret > 0) + ret *= static_cast( + config_.num_channels); // Return total number of samples. + *speech_type = ConvertSpeechType(temp_type); + return ret; +} + +int AudioDecoderMultiChannelOpusImpl::DecodeRedundantInternal( + const uint8_t* encoded, + size_t encoded_len, + int sample_rate_hz, + int16_t* decoded, + SpeechType* speech_type) { + if (!PacketHasFec(encoded, encoded_len)) { + // This packet is a RED packet. + return DecodeInternal(encoded, encoded_len, sample_rate_hz, decoded, + speech_type); + } + + RTC_DCHECK_EQ(sample_rate_hz, 48000); + int16_t temp_type = 1; // Default is speech. + int ret = WebRtcOpus_DecodeFec(dec_state_, encoded, encoded_len, decoded, + &temp_type); + if (ret > 0) + ret *= static_cast( + config_.num_channels); // Return total number of samples. + *speech_type = ConvertSpeechType(temp_type); + return ret; +} + +void AudioDecoderMultiChannelOpusImpl::Reset() { + WebRtcOpus_DecoderInit(dec_state_); +} + +int AudioDecoderMultiChannelOpusImpl::PacketDuration(const uint8_t* encoded, + size_t encoded_len) const { + return WebRtcOpus_DurationEst(dec_state_, encoded, encoded_len); +} + +int AudioDecoderMultiChannelOpusImpl::PacketDurationRedundant( + const uint8_t* encoded, + size_t encoded_len) const { + if (!PacketHasFec(encoded, encoded_len)) { + // This packet is a RED packet. + return PacketDuration(encoded, encoded_len); + } + + return WebRtcOpus_FecDurationEst(encoded, encoded_len, 48000); +} + +bool AudioDecoderMultiChannelOpusImpl::PacketHasFec(const uint8_t* encoded, + size_t encoded_len) const { + int fec; + fec = WebRtcOpus_PacketHasFec(encoded, encoded_len); + return (fec == 1); +} + +int AudioDecoderMultiChannelOpusImpl::SampleRateHz() const { + return 48000; +} + +size_t AudioDecoderMultiChannelOpusImpl::Channels() const { + return config_.num_channels; +} + +} // namespace webrtc diff --git a/third_party/libwebrtc/modules/audio_coding/codecs/opus/audio_decoder_multi_channel_opus_impl.h b/third_party/libwebrtc/modules/audio_coding/codecs/opus/audio_decoder_multi_channel_opus_impl.h new file mode 100644 index 0000000000..2ff47a8a53 --- /dev/null +++ b/third_party/libwebrtc/modules/audio_coding/codecs/opus/audio_decoder_multi_channel_opus_impl.h @@ -0,0 +1,74 @@ +/* + * Copyright (c) 2019 The WebRTC project authors. All Rights Reserved. + * + * Use of this source code is governed by a BSD-style license + * that can be found in the LICENSE file in the root of the source + * tree. An additional intellectual property rights grant can be found + * in the file PATENTS. All contributing project authors may + * be found in the AUTHORS file in the root of the source tree. + */ + +#ifndef MODULES_AUDIO_CODING_CODECS_OPUS_AUDIO_DECODER_MULTI_CHANNEL_OPUS_IMPL_H_ +#define MODULES_AUDIO_CODING_CODECS_OPUS_AUDIO_DECODER_MULTI_CHANNEL_OPUS_IMPL_H_ + +#include + +#include +#include + +#include "api/audio_codecs/audio_decoder.h" +#include "api/audio_codecs/audio_format.h" +#include "api/audio_codecs/opus/audio_decoder_multi_channel_opus_config.h" +#include "modules/audio_coding/codecs/opus/opus_interface.h" +#include "rtc_base/buffer.h" + +namespace webrtc { + +class AudioDecoderMultiChannelOpusImpl final : public AudioDecoder { + public: + static std::unique_ptr MakeAudioDecoder( + AudioDecoderMultiChannelOpusConfig config); + + ~AudioDecoderMultiChannelOpusImpl() override; + + AudioDecoderMultiChannelOpusImpl(const AudioDecoderMultiChannelOpusImpl&) = + delete; + AudioDecoderMultiChannelOpusImpl& operator=( + const AudioDecoderMultiChannelOpusImpl&) = delete; + + std::vector ParsePayload(rtc::Buffer&& payload, + uint32_t timestamp) override; + void Reset() override; + int PacketDuration(const uint8_t* encoded, size_t encoded_len) const override; + int PacketDurationRedundant(const uint8_t* encoded, + size_t encoded_len) const override; + bool PacketHasFec(const uint8_t* encoded, size_t encoded_len) const override; + int SampleRateHz() const override; + size_t Channels() const override; + + static absl::optional SdpToConfig( + const SdpAudioFormat& format); + + protected: + int DecodeInternal(const uint8_t* encoded, + size_t encoded_len, + int sample_rate_hz, + int16_t* decoded, + SpeechType* speech_type) override; + int DecodeRedundantInternal(const uint8_t* encoded, + size_t encoded_len, + int sample_rate_hz, + int16_t* decoded, + SpeechType* speech_type) override; + + private: + AudioDecoderMultiChannelOpusImpl(OpusDecInst* dec_state, + AudioDecoderMultiChannelOpusConfig config); + + OpusDecInst* dec_state_; + const AudioDecoderMultiChannelOpusConfig config_; +}; + +} // namespace webrtc + +#endif // MODULES_AUDIO_CODING_CODECS_OPUS_AUDIO_DECODER_MULTI_CHANNEL_OPUS_IMPL_H_ diff --git a/third_party/libwebrtc/modules/audio_coding/codecs/opus/audio_decoder_multi_channel_opus_unittest.cc b/third_party/libwebrtc/modules/audio_coding/codecs/opus/audio_decoder_multi_channel_opus_unittest.cc new file mode 100644 index 0000000000..57e2107f3c --- /dev/null +++ b/third_party/libwebrtc/modules/audio_coding/codecs/opus/audio_decoder_multi_channel_opus_unittest.cc @@ -0,0 +1,148 @@ +/* + * Copyright (c) 2019 The WebRTC project authors. All Rights Reserved. + * + * Use of this source code is governed by a BSD-style license + * that can be found in the LICENSE file in the root of the source + * tree. An additional intellectual property rights grant can be found + * in the file PATENTS. All contributing project authors may + * be found in the AUTHORS file in the root of the source tree. + */ + +#include "api/audio_codecs/opus/audio_decoder_multi_channel_opus.h" + +#include "modules/audio_coding/codecs/opus/audio_coder_opus_common.h" +#include "test/gmock.h" +#include "test/gtest.h" + +namespace webrtc { +using ::testing::NiceMock; +using ::testing::Return; + +TEST(AudioDecoderMultiOpusTest, GetFormatParameter) { + const SdpAudioFormat sdp_format("multiopus", 48000, 4, + {{"channel_mapping", "0,1,2,3"}, + {"coupled_streams", "2"}, + {"num_streams", "2"}}); + + EXPECT_EQ(GetFormatParameter(sdp_format, "channel_mapping"), + absl::optional("0,1,2,3")); + + EXPECT_EQ(GetFormatParameter(sdp_format, "coupled_streams"), + absl::optional(2)); + + EXPECT_EQ(GetFormatParameter(sdp_format, "missing"), absl::nullopt); + + EXPECT_EQ(GetFormatParameter(sdp_format, "channel_mapping"), + absl::nullopt); +} + +TEST(AudioDecoderMultiOpusTest, InvalidChannelMappings) { + { + // Can't use channel 3 if there are only 2 channels. + const SdpAudioFormat sdp_format("multiopus", 48000, 2, + {{"channel_mapping", "3,0"}, + {"coupled_streams", "1"}, + {"num_streams", "2"}}); + const absl::optional decoder_config = + AudioDecoderMultiChannelOpus::SdpToConfig(sdp_format); + EXPECT_FALSE(decoder_config.has_value()); + } + { + // The mapping is too long. There are only 5 channels, but 6 elements in the + // mapping. + const SdpAudioFormat sdp_format("multiopus", 48000, 5, + {{"channel_mapping", "0,1,2,3,4,5"}, + {"coupled_streams", "0"}, + {"num_streams", "2"}}); + const absl::optional decoder_config = + AudioDecoderMultiChannelOpus::SdpToConfig(sdp_format); + EXPECT_FALSE(decoder_config.has_value()); + } + { + // The mapping doesn't parse correctly. + const SdpAudioFormat sdp_format( + "multiopus", 48000, 5, + {{"channel_mapping", "0,1,two,3,4"}, {"coupled_streams", "0"}}); + const absl::optional decoder_config = + AudioDecoderMultiChannelOpus::SdpToConfig(sdp_format); + EXPECT_FALSE(decoder_config.has_value()); + } +} + +TEST(AudioDecoderMultiOpusTest, ValidSdpToConfigProducesCorrectConfig) { + const SdpAudioFormat sdp_format("multiopus", 48000, 4, + {{"channel_mapping", "3,1,2,0"}, + {"coupled_streams", "2"}, + {"num_streams", "2"}}); + + const absl::optional decoder_config = + AudioDecoderMultiChannelOpus::SdpToConfig(sdp_format); + + ASSERT_TRUE(decoder_config.has_value()); + EXPECT_TRUE(decoder_config->IsOk()); + EXPECT_EQ(decoder_config->coupled_streams, 2); + EXPECT_THAT(decoder_config->channel_mapping, + ::testing::ContainerEq(std::vector({3, 1, 2, 0}))); +} + +TEST(AudioDecoderMultiOpusTest, InvalidSdpToConfigDoesNotProduceConfig) { + { + const SdpAudioFormat sdp_format("multiopus", 48000, 4, + {{"channel_mapping", "0,1,2,3"}, + {"coupled_stream", "2"}, + {"num_streams", "2"}}); + + const absl::optional decoder_config = + AudioDecoderMultiChannelOpus::SdpToConfig(sdp_format); + + EXPECT_FALSE(decoder_config.has_value()); + } + + { + const SdpAudioFormat sdp_format("multiopus", 48000, 4, + {{"channel_mapping", "0,1,2 3"}, + {"coupled_streams", "2"}, + {"num_streams", "2"}}); + + const absl::optional decoder_config = + AudioDecoderMultiChannelOpus::SdpToConfig(sdp_format); + + EXPECT_FALSE(decoder_config.has_value()); + } +} + +TEST(AudioDecoderMultiOpusTest, CodecsCanBeCreated) { + const SdpAudioFormat sdp_format("multiopus", 48000, 4, + {{"channel_mapping", "0,1,2,3"}, + {"coupled_streams", "2"}, + {"num_streams", "2"}}); + + const absl::optional decoder_config = + AudioDecoderMultiChannelOpus::SdpToConfig(sdp_format); + + ASSERT_TRUE(decoder_config.has_value()); + + const std::unique_ptr opus_decoder = + AudioDecoderMultiChannelOpus::MakeAudioDecoder(*decoder_config); + + EXPECT_TRUE(opus_decoder); +} + +TEST(AudioDecoderMultiOpusTest, AdvertisedCodecsCanBeCreated) { + std::vector specs; + AudioDecoderMultiChannelOpus::AppendSupportedDecoders(&specs); + + EXPECT_FALSE(specs.empty()); + + for (const AudioCodecSpec& spec : specs) { + const absl::optional decoder_config = + AudioDecoderMultiChannelOpus::SdpToConfig(spec.format); + ASSERT_TRUE(decoder_config.has_value()); + + const std::unique_ptr opus_decoder = + AudioDecoderMultiChannelOpus::MakeAudioDecoder(*decoder_config); + + EXPECT_TRUE(opus_decoder); + } +} +} // namespace webrtc diff --git a/third_party/libwebrtc/modules/audio_coding/codecs/opus/audio_decoder_opus.cc b/third_party/libwebrtc/modules/audio_coding/codecs/opus/audio_decoder_opus.cc new file mode 100644 index 0000000000..cff9685548 --- /dev/null +++ b/third_party/libwebrtc/modules/audio_coding/codecs/opus/audio_decoder_opus.cc @@ -0,0 +1,128 @@ +/* + * Copyright (c) 2015 The WebRTC project authors. All Rights Reserved. + * + * Use of this source code is governed by a BSD-style license + * that can be found in the LICENSE file in the root of the source + * tree. An additional intellectual property rights grant can be found + * in the file PATENTS. All contributing project authors may + * be found in the AUTHORS file in the root of the source tree. + */ + +#include "modules/audio_coding/codecs/opus/audio_decoder_opus.h" + +#include +#include + +#include "absl/types/optional.h" +#include "api/array_view.h" +#include "modules/audio_coding/codecs/opus/audio_coder_opus_common.h" +#include "rtc_base/checks.h" + +namespace webrtc { + +AudioDecoderOpusImpl::AudioDecoderOpusImpl(size_t num_channels, + int sample_rate_hz) + : channels_{num_channels}, sample_rate_hz_{sample_rate_hz} { + RTC_DCHECK(num_channels == 1 || num_channels == 2); + RTC_DCHECK(sample_rate_hz == 16000 || sample_rate_hz == 48000); + const int error = + WebRtcOpus_DecoderCreate(&dec_state_, channels_, sample_rate_hz_); + RTC_DCHECK(error == 0); + WebRtcOpus_DecoderInit(dec_state_); +} + +AudioDecoderOpusImpl::~AudioDecoderOpusImpl() { + WebRtcOpus_DecoderFree(dec_state_); +} + +std::vector AudioDecoderOpusImpl::ParsePayload( + rtc::Buffer&& payload, + uint32_t timestamp) { + std::vector results; + + if (PacketHasFec(payload.data(), payload.size())) { + const int duration = + PacketDurationRedundant(payload.data(), payload.size()); + RTC_DCHECK_GE(duration, 0); + rtc::Buffer payload_copy(payload.data(), payload.size()); + std::unique_ptr fec_frame( + new OpusFrame(this, std::move(payload_copy), false)); + results.emplace_back(timestamp - duration, 1, std::move(fec_frame)); + } + std::unique_ptr frame( + new OpusFrame(this, std::move(payload), true)); + results.emplace_back(timestamp, 0, std::move(frame)); + return results; +} + +int AudioDecoderOpusImpl::DecodeInternal(const uint8_t* encoded, + size_t encoded_len, + int sample_rate_hz, + int16_t* decoded, + SpeechType* speech_type) { + RTC_DCHECK_EQ(sample_rate_hz, sample_rate_hz_); + int16_t temp_type = 1; // Default is speech. + int ret = + WebRtcOpus_Decode(dec_state_, encoded, encoded_len, decoded, &temp_type); + if (ret > 0) + ret *= static_cast(channels_); // Return total number of samples. + *speech_type = ConvertSpeechType(temp_type); + return ret; +} + +int AudioDecoderOpusImpl::DecodeRedundantInternal(const uint8_t* encoded, + size_t encoded_len, + int sample_rate_hz, + int16_t* decoded, + SpeechType* speech_type) { + if (!PacketHasFec(encoded, encoded_len)) { + // This packet is a RED packet. + return DecodeInternal(encoded, encoded_len, sample_rate_hz, decoded, + speech_type); + } + + RTC_DCHECK_EQ(sample_rate_hz, sample_rate_hz_); + int16_t temp_type = 1; // Default is speech. + int ret = WebRtcOpus_DecodeFec(dec_state_, encoded, encoded_len, decoded, + &temp_type); + if (ret > 0) + ret *= static_cast(channels_); // Return total number of samples. + *speech_type = ConvertSpeechType(temp_type); + return ret; +} + +void AudioDecoderOpusImpl::Reset() { + WebRtcOpus_DecoderInit(dec_state_); +} + +int AudioDecoderOpusImpl::PacketDuration(const uint8_t* encoded, + size_t encoded_len) const { + return WebRtcOpus_DurationEst(dec_state_, encoded, encoded_len); +} + +int AudioDecoderOpusImpl::PacketDurationRedundant(const uint8_t* encoded, + size_t encoded_len) const { + if (!PacketHasFec(encoded, encoded_len)) { + // This packet is a RED packet. + return PacketDuration(encoded, encoded_len); + } + + return WebRtcOpus_FecDurationEst(encoded, encoded_len, sample_rate_hz_); +} + +bool AudioDecoderOpusImpl::PacketHasFec(const uint8_t* encoded, + size_t encoded_len) const { + int fec; + fec = WebRtcOpus_PacketHasFec(encoded, encoded_len); + return (fec == 1); +} + +int AudioDecoderOpusImpl::SampleRateHz() const { + return sample_rate_hz_; +} + +size_t AudioDecoderOpusImpl::Channels() const { + return channels_; +} + +} // namespace webrtc diff --git a/third_party/libwebrtc/modules/audio_coding/codecs/opus/audio_decoder_opus.h b/third_party/libwebrtc/modules/audio_coding/codecs/opus/audio_decoder_opus.h new file mode 100644 index 0000000000..e8fd0440bc --- /dev/null +++ b/third_party/libwebrtc/modules/audio_coding/codecs/opus/audio_decoder_opus.h @@ -0,0 +1,64 @@ +/* + * Copyright (c) 2015 The WebRTC project authors. All Rights Reserved. + * + * Use of this source code is governed by a BSD-style license + * that can be found in the LICENSE file in the root of the source + * tree. An additional intellectual property rights grant can be found + * in the file PATENTS. All contributing project authors may + * be found in the AUTHORS file in the root of the source tree. + */ + +#ifndef MODULES_AUDIO_CODING_CODECS_OPUS_AUDIO_DECODER_OPUS_H_ +#define MODULES_AUDIO_CODING_CODECS_OPUS_AUDIO_DECODER_OPUS_H_ + +#include +#include + +#include + +#include "api/audio_codecs/audio_decoder.h" +#include "modules/audio_coding/codecs/opus/opus_interface.h" +#include "rtc_base/buffer.h" + +namespace webrtc { + +class AudioDecoderOpusImpl final : public AudioDecoder { + public: + explicit AudioDecoderOpusImpl(size_t num_channels, + int sample_rate_hz = 48000); + ~AudioDecoderOpusImpl() override; + + AudioDecoderOpusImpl(const AudioDecoderOpusImpl&) = delete; + AudioDecoderOpusImpl& operator=(const AudioDecoderOpusImpl&) = delete; + + std::vector ParsePayload(rtc::Buffer&& payload, + uint32_t timestamp) override; + void Reset() override; + int PacketDuration(const uint8_t* encoded, size_t encoded_len) const override; + int PacketDurationRedundant(const uint8_t* encoded, + size_t encoded_len) const override; + bool PacketHasFec(const uint8_t* encoded, size_t encoded_len) const override; + int SampleRateHz() const override; + size_t Channels() const override; + + protected: + int DecodeInternal(const uint8_t* encoded, + size_t encoded_len, + int sample_rate_hz, + int16_t* decoded, + SpeechType* speech_type) override; + int DecodeRedundantInternal(const uint8_t* encoded, + size_t encoded_len, + int sample_rate_hz, + int16_t* decoded, + SpeechType* speech_type) override; + + private: + OpusDecInst* dec_state_; + const size_t channels_; + const int sample_rate_hz_; +}; + +} // namespace webrtc + +#endif // MODULES_AUDIO_CODING_CODECS_OPUS_AUDIO_DECODER_OPUS_H_ diff --git a/third_party/libwebrtc/modules/audio_coding/codecs/opus/audio_encoder_multi_channel_opus_impl.cc b/third_party/libwebrtc/modules/audio_coding/codecs/opus/audio_encoder_multi_channel_opus_impl.cc new file mode 100644 index 0000000000..38a11c123d --- /dev/null +++ b/third_party/libwebrtc/modules/audio_coding/codecs/opus/audio_encoder_multi_channel_opus_impl.cc @@ -0,0 +1,366 @@ +/* + * Copyright (c) 2019 The WebRTC project authors. All Rights Reserved. + * + * Use of this source code is governed by a BSD-style license + * that can be found in the LICENSE file in the root of the source + * tree. An additional intellectual property rights grant can be found + * in the file PATENTS. All contributing project authors may + * be found in the AUTHORS file in the root of the source tree. + */ + +/* + * LEFT TO DO: + * - WRITE TESTS for the stuff in this file. + * - Check the creation, maybe make it safer by returning an empty optional or + * unique_ptr. --- It looks OK, but RecreateEncoderInstance can perhaps crash + * on a valid config. Can run it in the fuzzer for some time. Should prbl also + * fuzz the config. + */ + +#include "modules/audio_coding/codecs/opus/audio_encoder_multi_channel_opus_impl.h" + +#include +#include +#include +#include + +#include "absl/strings/match.h" +#include "modules/audio_coding/codecs/opus/audio_coder_opus_common.h" +#include "rtc_base/arraysize.h" +#include "rtc_base/checks.h" +#include "rtc_base/logging.h" +#include "rtc_base/string_to_number.h" + +namespace webrtc { + +namespace { + +// Recommended bitrates for one channel: +// 8-12 kb/s for NB speech, +// 16-20 kb/s for WB speech, +// 28-40 kb/s for FB speech, +// 48-64 kb/s for FB mono music, and +// 64-128 kb/s for FB stereo music. +// The current implementation multiplies these values by the number of channels. +constexpr int kOpusBitrateNbBps = 12000; +constexpr int kOpusBitrateWbBps = 20000; +constexpr int kOpusBitrateFbBps = 32000; + +constexpr int kDefaultMaxPlaybackRate = 48000; +// These two lists must be sorted from low to high +#if WEBRTC_OPUS_SUPPORT_120MS_PTIME +constexpr int kOpusSupportedFrameLengths[] = {10, 20, 40, 60, 120}; +#else +constexpr int kOpusSupportedFrameLengths[] = {10, 20, 40, 60}; +#endif + +int GetBitrateBps(const AudioEncoderMultiChannelOpusConfig& config) { + RTC_DCHECK(config.IsOk()); + return config.bitrate_bps; +} +int GetMaxPlaybackRate(const SdpAudioFormat& format) { + const auto param = GetFormatParameter(format, "maxplaybackrate"); + if (param && *param >= 8000) { + return std::min(*param, kDefaultMaxPlaybackRate); + } + return kDefaultMaxPlaybackRate; +} + +int GetFrameSizeMs(const SdpAudioFormat& format) { + const auto ptime = GetFormatParameter(format, "ptime"); + if (ptime.has_value()) { + // Pick the next highest supported frame length from + // kOpusSupportedFrameLengths. + for (const int supported_frame_length : kOpusSupportedFrameLengths) { + if (supported_frame_length >= *ptime) { + return supported_frame_length; + } + } + // If none was found, return the largest supported frame length. + return *(std::end(kOpusSupportedFrameLengths) - 1); + } + + return AudioEncoderOpusConfig::kDefaultFrameSizeMs; +} + +int CalculateDefaultBitrate(int max_playback_rate, size_t num_channels) { + const int bitrate = [&] { + if (max_playback_rate <= 8000) { + return kOpusBitrateNbBps * rtc::dchecked_cast(num_channels); + } else if (max_playback_rate <= 16000) { + return kOpusBitrateWbBps * rtc::dchecked_cast(num_channels); + } else { + return kOpusBitrateFbBps * rtc::dchecked_cast(num_channels); + } + }(); + RTC_DCHECK_GE(bitrate, AudioEncoderMultiChannelOpusConfig::kMinBitrateBps); + return bitrate; +} + +// Get the maxaveragebitrate parameter in string-form, so we can properly figure +// out how invalid it is and accurately log invalid values. +int CalculateBitrate(int max_playback_rate_hz, + size_t num_channels, + absl::optional bitrate_param) { + const int default_bitrate = + CalculateDefaultBitrate(max_playback_rate_hz, num_channels); + + if (bitrate_param) { + const auto bitrate = rtc::StringToNumber(*bitrate_param); + if (bitrate) { + const int chosen_bitrate = + std::max(AudioEncoderOpusConfig::kMinBitrateBps, + std::min(*bitrate, AudioEncoderOpusConfig::kMaxBitrateBps)); + if (bitrate != chosen_bitrate) { + RTC_LOG(LS_WARNING) << "Invalid maxaveragebitrate " << *bitrate + << " clamped to " << chosen_bitrate; + } + return chosen_bitrate; + } + RTC_LOG(LS_WARNING) << "Invalid maxaveragebitrate \"" << *bitrate_param + << "\" replaced by default bitrate " << default_bitrate; + } + + return default_bitrate; +} + +} // namespace + +std::unique_ptr +AudioEncoderMultiChannelOpusImpl::MakeAudioEncoder( + const AudioEncoderMultiChannelOpusConfig& config, + int payload_type) { + if (!config.IsOk()) { + RTC_DCHECK_NOTREACHED(); + return nullptr; + } + return std::make_unique(config, + payload_type); +} + +AudioEncoderMultiChannelOpusImpl::AudioEncoderMultiChannelOpusImpl( + const AudioEncoderMultiChannelOpusConfig& config, + int payload_type) + : payload_type_(payload_type), inst_(nullptr) { + RTC_DCHECK(0 <= payload_type && payload_type <= 127); + + RTC_CHECK(RecreateEncoderInstance(config)); +} + +AudioEncoderMultiChannelOpusImpl::~AudioEncoderMultiChannelOpusImpl() { + RTC_CHECK_EQ(0, WebRtcOpus_EncoderFree(inst_)); +} + +size_t AudioEncoderMultiChannelOpusImpl::SufficientOutputBufferSize() const { + // Calculate the number of bytes we expect the encoder to produce, + // then multiply by two to give a wide margin for error. + const size_t bytes_per_millisecond = + static_cast(GetBitrateBps(config_) / (1000 * 8) + 1); + const size_t approx_encoded_bytes = + Num10msFramesPerPacket() * 10 * bytes_per_millisecond; + return 2 * approx_encoded_bytes; +} + +void AudioEncoderMultiChannelOpusImpl::Reset() { + RTC_CHECK(RecreateEncoderInstance(config_)); +} + +absl::optional> +AudioEncoderMultiChannelOpusImpl::GetFrameLengthRange() const { + return {{TimeDelta::Millis(config_.frame_size_ms), + TimeDelta::Millis(config_.frame_size_ms)}}; +} + +// If the given config is OK, recreate the Opus encoder instance with those +// settings, save the config, and return true. Otherwise, do nothing and return +// false. +bool AudioEncoderMultiChannelOpusImpl::RecreateEncoderInstance( + const AudioEncoderMultiChannelOpusConfig& config) { + if (!config.IsOk()) + return false; + config_ = config; + if (inst_) + RTC_CHECK_EQ(0, WebRtcOpus_EncoderFree(inst_)); + input_buffer_.clear(); + input_buffer_.reserve(Num10msFramesPerPacket() * SamplesPer10msFrame()); + RTC_CHECK_EQ( + 0, WebRtcOpus_MultistreamEncoderCreate( + &inst_, config.num_channels, + config.application == + AudioEncoderMultiChannelOpusConfig::ApplicationMode::kVoip + ? 0 + : 1, + config.num_streams, config.coupled_streams, + config.channel_mapping.data())); + const int bitrate = GetBitrateBps(config); + RTC_CHECK_EQ(0, WebRtcOpus_SetBitRate(inst_, bitrate)); + RTC_LOG(LS_VERBOSE) << "Set Opus bitrate to " << bitrate << " bps."; + if (config.fec_enabled) { + RTC_CHECK_EQ(0, WebRtcOpus_EnableFec(inst_)); + RTC_LOG(LS_VERBOSE) << "Opus enable FEC"; + } else { + RTC_CHECK_EQ(0, WebRtcOpus_DisableFec(inst_)); + RTC_LOG(LS_VERBOSE) << "Opus disable FEC"; + } + RTC_CHECK_EQ( + 0, WebRtcOpus_SetMaxPlaybackRate(inst_, config.max_playback_rate_hz)); + RTC_LOG(LS_VERBOSE) << "Set Opus playback rate to " + << config.max_playback_rate_hz << " hz."; + + // Use the DEFAULT complexity. + RTC_CHECK_EQ( + 0, WebRtcOpus_SetComplexity(inst_, AudioEncoderOpusConfig().complexity)); + RTC_LOG(LS_VERBOSE) << "Set Opus coding complexity to " + << AudioEncoderOpusConfig().complexity; + + if (config.dtx_enabled) { + RTC_CHECK_EQ(0, WebRtcOpus_EnableDtx(inst_)); + RTC_LOG(LS_VERBOSE) << "Opus enable DTX"; + } else { + RTC_CHECK_EQ(0, WebRtcOpus_DisableDtx(inst_)); + RTC_LOG(LS_VERBOSE) << "Opus disable DTX"; + } + + if (config.cbr_enabled) { + RTC_CHECK_EQ(0, WebRtcOpus_EnableCbr(inst_)); + RTC_LOG(LS_VERBOSE) << "Opus enable CBR"; + } else { + RTC_CHECK_EQ(0, WebRtcOpus_DisableCbr(inst_)); + RTC_LOG(LS_VERBOSE) << "Opus disable CBR"; + } + num_channels_to_encode_ = NumChannels(); + next_frame_length_ms_ = config_.frame_size_ms; + RTC_LOG(LS_VERBOSE) << "Set Opus frame length to " << config_.frame_size_ms + << " ms"; + return true; +} + +absl::optional +AudioEncoderMultiChannelOpusImpl::SdpToConfig(const SdpAudioFormat& format) { + if (!absl::EqualsIgnoreCase(format.name, "multiopus") || + format.clockrate_hz != 48000) { + return absl::nullopt; + } + + AudioEncoderMultiChannelOpusConfig config; + config.num_channels = format.num_channels; + config.frame_size_ms = GetFrameSizeMs(format); + config.max_playback_rate_hz = GetMaxPlaybackRate(format); + config.fec_enabled = (GetFormatParameter(format, "useinbandfec") == "1"); + config.dtx_enabled = (GetFormatParameter(format, "usedtx") == "1"); + config.cbr_enabled = (GetFormatParameter(format, "cbr") == "1"); + config.bitrate_bps = + CalculateBitrate(config.max_playback_rate_hz, config.num_channels, + GetFormatParameter(format, "maxaveragebitrate")); + config.application = + config.num_channels == 1 + ? AudioEncoderMultiChannelOpusConfig::ApplicationMode::kVoip + : AudioEncoderMultiChannelOpusConfig::ApplicationMode::kAudio; + + config.supported_frame_lengths_ms.clear(); + std::copy(std::begin(kOpusSupportedFrameLengths), + std::end(kOpusSupportedFrameLengths), + std::back_inserter(config.supported_frame_lengths_ms)); + + auto num_streams = GetFormatParameter(format, "num_streams"); + if (!num_streams.has_value()) { + return absl::nullopt; + } + config.num_streams = *num_streams; + + auto coupled_streams = GetFormatParameter(format, "coupled_streams"); + if (!coupled_streams.has_value()) { + return absl::nullopt; + } + config.coupled_streams = *coupled_streams; + + auto channel_mapping = + GetFormatParameter>(format, "channel_mapping"); + if (!channel_mapping.has_value()) { + return absl::nullopt; + } + config.channel_mapping = *channel_mapping; + + if (!config.IsOk()) { + return absl::nullopt; + } + return config; +} + +AudioCodecInfo AudioEncoderMultiChannelOpusImpl::QueryAudioEncoder( + const AudioEncoderMultiChannelOpusConfig& config) { + RTC_DCHECK(config.IsOk()); + AudioCodecInfo info(48000, config.num_channels, config.bitrate_bps, + AudioEncoderOpusConfig::kMinBitrateBps, + AudioEncoderOpusConfig::kMaxBitrateBps); + info.allow_comfort_noise = false; + info.supports_network_adaption = false; + return info; +} + +size_t AudioEncoderMultiChannelOpusImpl::Num10msFramesPerPacket() const { + return static_cast(rtc::CheckedDivExact(config_.frame_size_ms, 10)); +} +size_t AudioEncoderMultiChannelOpusImpl::SamplesPer10msFrame() const { + return rtc::CheckedDivExact(48000, 100) * config_.num_channels; +} +int AudioEncoderMultiChannelOpusImpl::SampleRateHz() const { + return 48000; +} +size_t AudioEncoderMultiChannelOpusImpl::NumChannels() const { + return config_.num_channels; +} +size_t AudioEncoderMultiChannelOpusImpl::Num10MsFramesInNextPacket() const { + return Num10msFramesPerPacket(); +} +size_t AudioEncoderMultiChannelOpusImpl::Max10MsFramesInAPacket() const { + return Num10msFramesPerPacket(); +} +int AudioEncoderMultiChannelOpusImpl::GetTargetBitrate() const { + return GetBitrateBps(config_); +} + +AudioEncoder::EncodedInfo AudioEncoderMultiChannelOpusImpl::EncodeImpl( + uint32_t rtp_timestamp, + rtc::ArrayView audio, + rtc::Buffer* encoded) { + if (input_buffer_.empty()) + first_timestamp_in_buffer_ = rtp_timestamp; + + input_buffer_.insert(input_buffer_.end(), audio.cbegin(), audio.cend()); + if (input_buffer_.size() < + (Num10msFramesPerPacket() * SamplesPer10msFrame())) { + return EncodedInfo(); + } + RTC_CHECK_EQ(input_buffer_.size(), + Num10msFramesPerPacket() * SamplesPer10msFrame()); + + const size_t max_encoded_bytes = SufficientOutputBufferSize(); + EncodedInfo info; + info.encoded_bytes = encoded->AppendData( + max_encoded_bytes, [&](rtc::ArrayView encoded) { + int status = WebRtcOpus_Encode( + inst_, &input_buffer_[0], + rtc::CheckedDivExact(input_buffer_.size(), config_.num_channels), + rtc::saturated_cast(max_encoded_bytes), encoded.data()); + + RTC_CHECK_GE(status, 0); // Fails only if fed invalid data. + + return static_cast(status); + }); + input_buffer_.clear(); + + // Will use new packet size for next encoding. + config_.frame_size_ms = next_frame_length_ms_; + + info.encoded_timestamp = first_timestamp_in_buffer_; + info.payload_type = payload_type_; + info.send_even_if_empty = true; // Allows Opus to send empty packets. + + info.speech = true; + info.encoder_type = CodecType::kOther; + + return info; +} + +} // namespace webrtc diff --git a/third_party/libwebrtc/modules/audio_coding/codecs/opus/audio_encoder_multi_channel_opus_impl.h b/third_party/libwebrtc/modules/audio_coding/codecs/opus/audio_encoder_multi_channel_opus_impl.h new file mode 100644 index 0000000000..8a7210515c --- /dev/null +++ b/third_party/libwebrtc/modules/audio_coding/codecs/opus/audio_encoder_multi_channel_opus_impl.h @@ -0,0 +1,92 @@ +/* + * Copyright (c) 2019 The WebRTC project authors. All Rights Reserved. + * + * Use of this source code is governed by a BSD-style license + * that can be found in the LICENSE file in the root of the source + * tree. An additional intellectual property rights grant can be found + * in the file PATENTS. All contributing project authors may + * be found in the AUTHORS file in the root of the source tree. + */ + +#ifndef MODULES_AUDIO_CODING_CODECS_OPUS_AUDIO_ENCODER_MULTI_CHANNEL_OPUS_IMPL_H_ +#define MODULES_AUDIO_CODING_CODECS_OPUS_AUDIO_ENCODER_MULTI_CHANNEL_OPUS_IMPL_H_ + +#include +#include +#include + +#include "absl/types/optional.h" +#include "api/audio_codecs/audio_encoder.h" +#include "api/audio_codecs/audio_format.h" +#include "api/audio_codecs/opus/audio_encoder_multi_channel_opus_config.h" +#include "api/units/time_delta.h" +#include "modules/audio_coding/codecs/opus/opus_interface.h" + +namespace webrtc { + +class RtcEventLog; + +class AudioEncoderMultiChannelOpusImpl final : public AudioEncoder { + public: + AudioEncoderMultiChannelOpusImpl( + const AudioEncoderMultiChannelOpusConfig& config, + int payload_type); + ~AudioEncoderMultiChannelOpusImpl() override; + + AudioEncoderMultiChannelOpusImpl(const AudioEncoderMultiChannelOpusImpl&) = + delete; + AudioEncoderMultiChannelOpusImpl& operator=( + const AudioEncoderMultiChannelOpusImpl&) = delete; + + // Static interface for use by BuiltinAudioEncoderFactory. + static constexpr const char* GetPayloadName() { return "multiopus"; } + static absl::optional QueryAudioEncoder( + const SdpAudioFormat& format); + + int SampleRateHz() const override; + size_t NumChannels() const override; + size_t Num10MsFramesInNextPacket() const override; + size_t Max10MsFramesInAPacket() const override; + int GetTargetBitrate() const override; + + void Reset() override; + absl::optional> GetFrameLengthRange() + const override; + + protected: + EncodedInfo EncodeImpl(uint32_t rtp_timestamp, + rtc::ArrayView audio, + rtc::Buffer* encoded) override; + + private: + static absl::optional SdpToConfig( + const SdpAudioFormat& format); + static AudioCodecInfo QueryAudioEncoder( + const AudioEncoderMultiChannelOpusConfig& config); + static std::unique_ptr MakeAudioEncoder( + const AudioEncoderMultiChannelOpusConfig&, + int payload_type); + + size_t Num10msFramesPerPacket() const; + size_t SamplesPer10msFrame() const; + size_t SufficientOutputBufferSize() const; + bool RecreateEncoderInstance( + const AudioEncoderMultiChannelOpusConfig& config); + void SetFrameLength(int frame_length_ms); + void SetNumChannelsToEncode(size_t num_channels_to_encode); + void SetProjectedPacketLossRate(float fraction); + + AudioEncoderMultiChannelOpusConfig config_; + const int payload_type_; + std::vector input_buffer_; + OpusEncInst* inst_; + uint32_t first_timestamp_in_buffer_; + size_t num_channels_to_encode_; + int next_frame_length_ms_; + + friend struct AudioEncoderMultiChannelOpus; +}; + +} // namespace webrtc + +#endif // MODULES_AUDIO_CODING_CODECS_OPUS_AUDIO_ENCODER_MULTI_CHANNEL_OPUS_IMPL_H_ diff --git a/third_party/libwebrtc/modules/audio_coding/codecs/opus/audio_encoder_multi_channel_opus_unittest.cc b/third_party/libwebrtc/modules/audio_coding/codecs/opus/audio_encoder_multi_channel_opus_unittest.cc new file mode 100644 index 0000000000..92f6f2c169 --- /dev/null +++ b/third_party/libwebrtc/modules/audio_coding/codecs/opus/audio_encoder_multi_channel_opus_unittest.cc @@ -0,0 +1,156 @@ +/* + * Copyright (c) 2019 The WebRTC project authors. All Rights Reserved. + * + * Use of this source code is governed by a BSD-style license + * that can be found in the LICENSE file in the root of the source + * tree. An additional intellectual property rights grant can be found + * in the file PATENTS. All contributing project authors may + * be found in the AUTHORS file in the root of the source tree. + */ + +#include "api/audio_codecs/opus/audio_encoder_multi_channel_opus.h" + +#include "test/gmock.h" + +namespace webrtc { +using ::testing::NiceMock; +using ::testing::Return; + +namespace { +constexpr int kOpusPayloadType = 120; +} // namespace + +TEST(AudioEncoderMultiOpusTest, CheckConfigValidity) { + { + const SdpAudioFormat sdp_format("multiopus", 48000, 2, + {{"channel_mapping", "3,0"}, + {"coupled_streams", "1"}, + {"num_streams", "2"}}); + const absl::optional encoder_config = + AudioEncoderMultiChannelOpus::SdpToConfig(sdp_format); + + // Maps input channel 0 to coded channel 3, which doesn't exist. + EXPECT_FALSE(encoder_config.has_value()); + } + + { + const SdpAudioFormat sdp_format("multiopus", 48000, 2, + {{"channel_mapping", "0"}, + {"coupled_streams", "1"}, + {"num_streams", "2"}}); + const absl::optional encoder_config = + AudioEncoderMultiChannelOpus::SdpToConfig(sdp_format); + + // The mapping is too short. + EXPECT_FALSE(encoder_config.has_value()); + } + { + const SdpAudioFormat sdp_format("multiopus", 48000, 3, + {{"channel_mapping", "0,0,0"}, + {"coupled_streams", "0"}, + {"num_streams", "1"}}); + const absl::optional encoder_config = + AudioEncoderMultiChannelOpus::SdpToConfig(sdp_format); + + // Coded channel 0 comes from both input channels 0, 1 and 2. + EXPECT_FALSE(encoder_config.has_value()); + } + { + const SdpAudioFormat sdp_format("multiopus", 48000, 3, + {{"channel_mapping", "0,255,255"}, + {"coupled_streams", "0"}, + {"num_streams", "1"}}); + const absl::optional encoder_config = + AudioEncoderMultiChannelOpus::SdpToConfig(sdp_format); + ASSERT_TRUE(encoder_config.has_value()); + + // This is fine, because channels 1, 2 are set to be ignored. + EXPECT_TRUE(encoder_config->IsOk()); + } + { + const SdpAudioFormat sdp_format("multiopus", 48000, 3, + {{"channel_mapping", "0,255,255"}, + {"coupled_streams", "0"}, + {"num_streams", "2"}}); + const absl::optional encoder_config = + AudioEncoderMultiChannelOpus::SdpToConfig(sdp_format); + + // This is NOT fine, because channels nothing says how coded channel 1 + // should be coded. + EXPECT_FALSE(encoder_config.has_value()); + } +} + +TEST(AudioEncoderMultiOpusTest, ConfigValuesAreParsedCorrectly) { + SdpAudioFormat sdp_format({"multiopus", + 48000, + 6, + {{"minptime", "10"}, + {"useinbandfec", "1"}, + {"channel_mapping", "0,4,1,2,3,5"}, + {"num_streams", "4"}, + {"coupled_streams", "2"}}}); + const absl::optional encoder_config = + AudioEncoderMultiChannelOpus::SdpToConfig(sdp_format); + ASSERT_TRUE(encoder_config.has_value()); + + EXPECT_EQ(encoder_config->coupled_streams, 2); + EXPECT_EQ(encoder_config->num_streams, 4); + EXPECT_THAT( + encoder_config->channel_mapping, + testing::ContainerEq(std::vector({0, 4, 1, 2, 3, 5}))); +} + +TEST(AudioEncoderMultiOpusTest, CreateFromValidConfig) { + { + const SdpAudioFormat sdp_format("multiopus", 48000, 3, + {{"channel_mapping", "0,255,255"}, + {"coupled_streams", "0"}, + {"num_streams", "2"}}); + const absl::optional encoder_config = + AudioEncoderMultiChannelOpus::SdpToConfig(sdp_format); + ASSERT_FALSE(encoder_config.has_value()); + } + { + const SdpAudioFormat sdp_format("multiopus", 48000, 3, + {{"channel_mapping", "1,255,0"}, + {"coupled_streams", "1"}, + {"num_streams", "1"}}); + const absl::optional encoder_config = + AudioEncoderMultiChannelOpus::SdpToConfig(sdp_format); + ASSERT_TRUE(encoder_config.has_value()); + + EXPECT_THAT(encoder_config->channel_mapping, + testing::ContainerEq(std::vector({1, 255, 0}))); + + EXPECT_TRUE(encoder_config->IsOk()); + + const std::unique_ptr opus_encoder = + AudioEncoderMultiChannelOpus::MakeAudioEncoder(*encoder_config, + kOpusPayloadType); + + // Creating an encoder from a valid config should work. + EXPECT_TRUE(opus_encoder); + } +} + +TEST(AudioEncoderMultiOpusTest, AdvertisedCodecsCanBeCreated) { + std::vector specs; + AudioEncoderMultiChannelOpus::AppendSupportedEncoders(&specs); + + EXPECT_FALSE(specs.empty()); + + for (const AudioCodecSpec& spec : specs) { + const absl::optional encoder_config = + AudioEncoderMultiChannelOpus::SdpToConfig(spec.format); + ASSERT_TRUE(encoder_config.has_value()); + + const std::unique_ptr opus_encoder = + AudioEncoderMultiChannelOpus::MakeAudioEncoder(*encoder_config, + kOpusPayloadType); + + EXPECT_TRUE(opus_encoder); + } +} + +} // namespace webrtc diff --git a/third_party/libwebrtc/modules/audio_coding/codecs/opus/audio_encoder_opus.cc b/third_party/libwebrtc/modules/audio_coding/codecs/opus/audio_encoder_opus.cc new file mode 100644 index 0000000000..17e0e33b1d --- /dev/null +++ b/third_party/libwebrtc/modules/audio_coding/codecs/opus/audio_encoder_opus.cc @@ -0,0 +1,824 @@ +/* + * Copyright (c) 2014 The WebRTC project authors. All Rights Reserved. + * + * Use of this source code is governed by a BSD-style license + * that can be found in the LICENSE file in the root of the source + * tree. An additional intellectual property rights grant can be found + * in the file PATENTS. All contributing project authors may + * be found in the AUTHORS file in the root of the source tree. + */ + +#include "modules/audio_coding/codecs/opus/audio_encoder_opus.h" + +#include +#include +#include +#include +#include + +#include "absl/strings/match.h" +#include "absl/strings/string_view.h" +#include "modules/audio_coding/audio_network_adaptor/audio_network_adaptor_impl.h" +#include "modules/audio_coding/audio_network_adaptor/controller_manager.h" +#include "modules/audio_coding/codecs/opus/audio_coder_opus_common.h" +#include "modules/audio_coding/codecs/opus/opus_interface.h" +#include "rtc_base/arraysize.h" +#include "rtc_base/checks.h" +#include "rtc_base/logging.h" +#include "rtc_base/numerics/exp_filter.h" +#include "rtc_base/numerics/safe_conversions.h" +#include "rtc_base/numerics/safe_minmax.h" +#include "rtc_base/string_encode.h" +#include "rtc_base/string_to_number.h" +#include "rtc_base/time_utils.h" +#include "system_wrappers/include/field_trial.h" + +namespace webrtc { + +namespace { + +// Codec parameters for Opus. +// draft-spittka-payload-rtp-opus-03 + +// Recommended bitrates: +// 8-12 kb/s for NB speech, +// 16-20 kb/s for WB speech, +// 28-40 kb/s for FB speech, +// 48-64 kb/s for FB mono music, and +// 64-128 kb/s for FB stereo music. +// The current implementation applies the following values to mono signals, +// and multiplies them by 2 for stereo. +constexpr int kOpusBitrateNbBps = 12000; +constexpr int kOpusBitrateWbBps = 20000; +constexpr int kOpusBitrateFbBps = 32000; + +constexpr int kRtpTimestampRateHz = 48000; +constexpr int kDefaultMaxPlaybackRate = 48000; + +// These two lists must be sorted from low to high +#if WEBRTC_OPUS_SUPPORT_120MS_PTIME +constexpr int kANASupportedFrameLengths[] = {20, 40, 60, 120}; +constexpr int kOpusSupportedFrameLengths[] = {10, 20, 40, 60, 120}; +#else +constexpr int kANASupportedFrameLengths[] = {20, 40, 60}; +constexpr int kOpusSupportedFrameLengths[] = {10, 20, 40, 60}; +#endif + +// PacketLossFractionSmoother uses an exponential filter with a time constant +// of -1.0 / ln(0.9999) = 10000 ms. +constexpr float kAlphaForPacketLossFractionSmoother = 0.9999f; +constexpr float kMaxPacketLossFraction = 0.2f; + +int CalculateDefaultBitrate(int max_playback_rate, size_t num_channels) { + const int bitrate = [&] { + if (max_playback_rate <= 8000) { + return kOpusBitrateNbBps * rtc::dchecked_cast(num_channels); + } else if (max_playback_rate <= 16000) { + return kOpusBitrateWbBps * rtc::dchecked_cast(num_channels); + } else { + return kOpusBitrateFbBps * rtc::dchecked_cast(num_channels); + } + }(); + RTC_DCHECK_GE(bitrate, AudioEncoderOpusConfig::kMinBitrateBps); + RTC_DCHECK_LE(bitrate, AudioEncoderOpusConfig::kMaxBitrateBps); + return bitrate; +} + +// Get the maxaveragebitrate parameter in string-form, so we can properly figure +// out how invalid it is and accurately log invalid values. +int CalculateBitrate(int max_playback_rate_hz, + size_t num_channels, + absl::optional bitrate_param) { + const int default_bitrate = + CalculateDefaultBitrate(max_playback_rate_hz, num_channels); + + if (bitrate_param) { + const auto bitrate = rtc::StringToNumber(*bitrate_param); + if (bitrate) { + const int chosen_bitrate = + std::max(AudioEncoderOpusConfig::kMinBitrateBps, + std::min(*bitrate, AudioEncoderOpusConfig::kMaxBitrateBps)); + if (bitrate != chosen_bitrate) { + RTC_LOG(LS_WARNING) << "Invalid maxaveragebitrate " << *bitrate + << " clamped to " << chosen_bitrate; + } + return chosen_bitrate; + } + RTC_LOG(LS_WARNING) << "Invalid maxaveragebitrate \"" << *bitrate_param + << "\" replaced by default bitrate " << default_bitrate; + } + + return default_bitrate; +} + +int GetChannelCount(const SdpAudioFormat& format) { + const auto param = GetFormatParameter(format, "stereo"); + if (param == "1") { + return 2; + } else { + return 1; + } +} + +int GetMaxPlaybackRate(const SdpAudioFormat& format) { + const auto param = GetFormatParameter(format, "maxplaybackrate"); + if (param && *param >= 8000) { + return std::min(*param, kDefaultMaxPlaybackRate); + } + return kDefaultMaxPlaybackRate; +} + +int GetFrameSizeMs(const SdpAudioFormat& format) { + const auto ptime = GetFormatParameter(format, "ptime"); + if (ptime) { + // Pick the next highest supported frame length from + // kOpusSupportedFrameLengths. + for (const int supported_frame_length : kOpusSupportedFrameLengths) { + if (supported_frame_length >= *ptime) { + return supported_frame_length; + } + } + // If none was found, return the largest supported frame length. + return *(std::end(kOpusSupportedFrameLengths) - 1); + } + + return AudioEncoderOpusConfig::kDefaultFrameSizeMs; +} + +void FindSupportedFrameLengths(int min_frame_length_ms, + int max_frame_length_ms, + std::vector* out) { + out->clear(); + std::copy_if(std::begin(kANASupportedFrameLengths), + std::end(kANASupportedFrameLengths), std::back_inserter(*out), + [&](int frame_length_ms) { + return frame_length_ms >= min_frame_length_ms && + frame_length_ms <= max_frame_length_ms; + }); + RTC_DCHECK(std::is_sorted(out->begin(), out->end())); +} + +int GetBitrateBps(const AudioEncoderOpusConfig& config) { + RTC_DCHECK(config.IsOk()); + return *config.bitrate_bps; +} + +std::vector GetBitrateMultipliers() { + constexpr char kBitrateMultipliersName[] = + "WebRTC-Audio-OpusBitrateMultipliers"; + const bool use_bitrate_multipliers = + webrtc::field_trial::IsEnabled(kBitrateMultipliersName); + if (use_bitrate_multipliers) { + const std::string field_trial_string = + webrtc::field_trial::FindFullName(kBitrateMultipliersName); + std::vector pieces; + rtc::tokenize(field_trial_string, '-', &pieces); + if (pieces.size() < 2 || pieces[0] != "Enabled") { + RTC_LOG(LS_WARNING) << "Invalid parameters for " + << kBitrateMultipliersName + << ", not using custom values."; + return std::vector(); + } + std::vector multipliers(pieces.size() - 1); + for (size_t i = 1; i < pieces.size(); i++) { + if (!rtc::FromString(pieces[i], &multipliers[i - 1])) { + RTC_LOG(LS_WARNING) + << "Invalid parameters for " << kBitrateMultipliersName + << ", not using custom values."; + return std::vector(); + } + } + RTC_LOG(LS_INFO) << "Using custom bitrate multipliers: " + << field_trial_string; + return multipliers; + } + return std::vector(); +} + +int GetMultipliedBitrate(int bitrate, const std::vector& multipliers) { + // The multipliers are valid from 5 kbps. + const size_t bitrate_kbps = static_cast(bitrate / 1000); + if (bitrate_kbps < 5 || bitrate_kbps >= multipliers.size() + 5) { + return bitrate; + } + return static_cast(multipliers[bitrate_kbps - 5] * bitrate); +} +} // namespace + +void AudioEncoderOpusImpl::AppendSupportedEncoders( + std::vector* specs) { + const SdpAudioFormat fmt = {"opus", + kRtpTimestampRateHz, + 2, + {{"minptime", "10"}, {"useinbandfec", "1"}}}; + const AudioCodecInfo info = QueryAudioEncoder(*SdpToConfig(fmt)); + specs->push_back({fmt, info}); +} + +AudioCodecInfo AudioEncoderOpusImpl::QueryAudioEncoder( + const AudioEncoderOpusConfig& config) { + RTC_DCHECK(config.IsOk()); + AudioCodecInfo info(config.sample_rate_hz, config.num_channels, + *config.bitrate_bps, + AudioEncoderOpusConfig::kMinBitrateBps, + AudioEncoderOpusConfig::kMaxBitrateBps); + info.allow_comfort_noise = false; + info.supports_network_adaption = true; + return info; +} + +std::unique_ptr AudioEncoderOpusImpl::MakeAudioEncoder( + const AudioEncoderOpusConfig& config, + int payload_type) { + if (!config.IsOk()) { + RTC_DCHECK_NOTREACHED(); + return nullptr; + } + return std::make_unique(config, payload_type); +} + +absl::optional AudioEncoderOpusImpl::SdpToConfig( + const SdpAudioFormat& format) { + if (!absl::EqualsIgnoreCase(format.name, "opus") || + format.clockrate_hz != kRtpTimestampRateHz) { + return absl::nullopt; + } + + AudioEncoderOpusConfig config; + config.num_channels = GetChannelCount(format); + config.frame_size_ms = GetFrameSizeMs(format); + config.max_playback_rate_hz = GetMaxPlaybackRate(format); + config.fec_enabled = (GetFormatParameter(format, "useinbandfec") == "1"); + config.dtx_enabled = (GetFormatParameter(format, "usedtx") == "1"); + config.cbr_enabled = (GetFormatParameter(format, "cbr") == "1"); + config.bitrate_bps = + CalculateBitrate(config.max_playback_rate_hz, config.num_channels, + GetFormatParameter(format, "maxaveragebitrate")); + config.application = config.num_channels == 1 + ? AudioEncoderOpusConfig::ApplicationMode::kVoip + : AudioEncoderOpusConfig::ApplicationMode::kAudio; + + constexpr int kMinANAFrameLength = kANASupportedFrameLengths[0]; + constexpr int kMaxANAFrameLength = + kANASupportedFrameLengths[arraysize(kANASupportedFrameLengths) - 1]; + + // For now, minptime and maxptime are only used with ANA. If ptime is outside + // of this range, it will get adjusted once ANA takes hold. Ideally, we'd know + // if ANA was to be used when setting up the config, and adjust accordingly. + const int min_frame_length_ms = + GetFormatParameter(format, "minptime").value_or(kMinANAFrameLength); + const int max_frame_length_ms = + GetFormatParameter(format, "maxptime").value_or(kMaxANAFrameLength); + + FindSupportedFrameLengths(min_frame_length_ms, max_frame_length_ms, + &config.supported_frame_lengths_ms); + if (!config.IsOk()) { + RTC_DCHECK_NOTREACHED(); + return absl::nullopt; + } + return config; +} + +absl::optional AudioEncoderOpusImpl::GetNewComplexity( + const AudioEncoderOpusConfig& config) { + RTC_DCHECK(config.IsOk()); + const int bitrate_bps = GetBitrateBps(config); + if (bitrate_bps >= config.complexity_threshold_bps - + config.complexity_threshold_window_bps && + bitrate_bps <= config.complexity_threshold_bps + + config.complexity_threshold_window_bps) { + // Within the hysteresis window; make no change. + return absl::nullopt; + } else { + return bitrate_bps <= config.complexity_threshold_bps + ? config.low_rate_complexity + : config.complexity; + } +} + +absl::optional AudioEncoderOpusImpl::GetNewBandwidth( + const AudioEncoderOpusConfig& config, + OpusEncInst* inst) { + constexpr int kMinWidebandBitrate = 8000; + constexpr int kMaxNarrowbandBitrate = 9000; + constexpr int kAutomaticThreshold = 11000; + RTC_DCHECK(config.IsOk()); + const int bitrate = GetBitrateBps(config); + if (bitrate > kAutomaticThreshold) { + return absl::optional(OPUS_AUTO); + } + const int bandwidth = WebRtcOpus_GetBandwidth(inst); + RTC_DCHECK_GE(bandwidth, 0); + if (bitrate > kMaxNarrowbandBitrate && bandwidth < OPUS_BANDWIDTH_WIDEBAND) { + return absl::optional(OPUS_BANDWIDTH_WIDEBAND); + } else if (bitrate < kMinWidebandBitrate && + bandwidth > OPUS_BANDWIDTH_NARROWBAND) { + return absl::optional(OPUS_BANDWIDTH_NARROWBAND); + } + return absl::optional(); +} + +class AudioEncoderOpusImpl::PacketLossFractionSmoother { + public: + explicit PacketLossFractionSmoother() + : last_sample_time_ms_(rtc::TimeMillis()), + smoother_(kAlphaForPacketLossFractionSmoother) {} + + // Gets the smoothed packet loss fraction. + float GetAverage() const { + float value = smoother_.filtered(); + return (value == rtc::ExpFilter::kValueUndefined) ? 0.0f : value; + } + + // Add new observation to the packet loss fraction smoother. + void AddSample(float packet_loss_fraction) { + int64_t now_ms = rtc::TimeMillis(); + smoother_.Apply(static_cast(now_ms - last_sample_time_ms_), + packet_loss_fraction); + last_sample_time_ms_ = now_ms; + } + + private: + int64_t last_sample_time_ms_; + + // An exponential filter is used to smooth the packet loss fraction. + rtc::ExpFilter smoother_; +}; + +AudioEncoderOpusImpl::AudioEncoderOpusImpl(const AudioEncoderOpusConfig& config, + int payload_type) + : AudioEncoderOpusImpl( + config, + payload_type, + [this](absl::string_view config_string, RtcEventLog* event_log) { + return DefaultAudioNetworkAdaptorCreator(config_string, event_log); + }, + // We choose 5sec as initial time constant due to empirical data. + std::make_unique(5000)) {} + +AudioEncoderOpusImpl::AudioEncoderOpusImpl( + const AudioEncoderOpusConfig& config, + int payload_type, + const AudioNetworkAdaptorCreator& audio_network_adaptor_creator, + std::unique_ptr bitrate_smoother) + : payload_type_(payload_type), + use_stable_target_for_adaptation_(!webrtc::field_trial::IsDisabled( + "WebRTC-Audio-StableTargetAdaptation")), + adjust_bandwidth_( + webrtc::field_trial::IsEnabled("WebRTC-AdjustOpusBandwidth")), + bitrate_changed_(true), + bitrate_multipliers_(GetBitrateMultipliers()), + packet_loss_rate_(0.0), + inst_(nullptr), + packet_loss_fraction_smoother_(new PacketLossFractionSmoother()), + audio_network_adaptor_creator_(audio_network_adaptor_creator), + bitrate_smoother_(std::move(bitrate_smoother)), + consecutive_dtx_frames_(0) { + RTC_DCHECK(0 <= payload_type && payload_type <= 127); + + // Sanity check of the redundant payload type field that we want to get rid + // of. See https://bugs.chromium.org/p/webrtc/issues/detail?id=7847 + RTC_CHECK(config.payload_type == -1 || config.payload_type == payload_type); + + RTC_CHECK(RecreateEncoderInstance(config)); + SetProjectedPacketLossRate(packet_loss_rate_); +} + +AudioEncoderOpusImpl::AudioEncoderOpusImpl(int payload_type, + const SdpAudioFormat& format) + : AudioEncoderOpusImpl(*SdpToConfig(format), payload_type) {} + +AudioEncoderOpusImpl::~AudioEncoderOpusImpl() { + RTC_CHECK_EQ(0, WebRtcOpus_EncoderFree(inst_)); +} + +int AudioEncoderOpusImpl::SampleRateHz() const { + return config_.sample_rate_hz; +} + +size_t AudioEncoderOpusImpl::NumChannels() const { + return config_.num_channels; +} + +int AudioEncoderOpusImpl::RtpTimestampRateHz() const { + return kRtpTimestampRateHz; +} + +size_t AudioEncoderOpusImpl::Num10MsFramesInNextPacket() const { + return Num10msFramesPerPacket(); +} + +size_t AudioEncoderOpusImpl::Max10MsFramesInAPacket() const { + return Num10msFramesPerPacket(); +} + +int AudioEncoderOpusImpl::GetTargetBitrate() const { + return GetBitrateBps(config_); +} + +void AudioEncoderOpusImpl::Reset() { + RTC_CHECK(RecreateEncoderInstance(config_)); +} + +bool AudioEncoderOpusImpl::SetFec(bool enable) { + if (enable) { + RTC_CHECK_EQ(0, WebRtcOpus_EnableFec(inst_)); + } else { + RTC_CHECK_EQ(0, WebRtcOpus_DisableFec(inst_)); + } + config_.fec_enabled = enable; + return true; +} + +bool AudioEncoderOpusImpl::SetDtx(bool enable) { + if (enable) { + RTC_CHECK_EQ(0, WebRtcOpus_EnableDtx(inst_)); + } else { + RTC_CHECK_EQ(0, WebRtcOpus_DisableDtx(inst_)); + } + config_.dtx_enabled = enable; + return true; +} + +bool AudioEncoderOpusImpl::GetDtx() const { + return config_.dtx_enabled; +} + +bool AudioEncoderOpusImpl::SetApplication(Application application) { + auto conf = config_; + switch (application) { + case Application::kSpeech: + conf.application = AudioEncoderOpusConfig::ApplicationMode::kVoip; + break; + case Application::kAudio: + conf.application = AudioEncoderOpusConfig::ApplicationMode::kAudio; + break; + } + return RecreateEncoderInstance(conf); +} + +void AudioEncoderOpusImpl::SetMaxPlaybackRate(int frequency_hz) { + auto conf = config_; + conf.max_playback_rate_hz = frequency_hz; + RTC_CHECK(RecreateEncoderInstance(conf)); +} + +bool AudioEncoderOpusImpl::EnableAudioNetworkAdaptor( + const std::string& config_string, + RtcEventLog* event_log) { + audio_network_adaptor_ = + audio_network_adaptor_creator_(config_string, event_log); + return audio_network_adaptor_.get() != nullptr; +} + +void AudioEncoderOpusImpl::DisableAudioNetworkAdaptor() { + audio_network_adaptor_.reset(nullptr); +} + +void AudioEncoderOpusImpl::OnReceivedUplinkPacketLossFraction( + float uplink_packet_loss_fraction) { + if (audio_network_adaptor_) { + audio_network_adaptor_->SetUplinkPacketLossFraction( + uplink_packet_loss_fraction); + ApplyAudioNetworkAdaptor(); + } + packet_loss_fraction_smoother_->AddSample(uplink_packet_loss_fraction); + float average_fraction_loss = packet_loss_fraction_smoother_->GetAverage(); + SetProjectedPacketLossRate(average_fraction_loss); +} + +void AudioEncoderOpusImpl::OnReceivedTargetAudioBitrate( + int target_audio_bitrate_bps) { + SetTargetBitrate(target_audio_bitrate_bps); +} + +void AudioEncoderOpusImpl::OnReceivedUplinkBandwidth( + int target_audio_bitrate_bps, + absl::optional bwe_period_ms, + absl::optional stable_target_bitrate_bps) { + if (audio_network_adaptor_) { + audio_network_adaptor_->SetTargetAudioBitrate(target_audio_bitrate_bps); + if (use_stable_target_for_adaptation_) { + if (stable_target_bitrate_bps) + audio_network_adaptor_->SetUplinkBandwidth(*stable_target_bitrate_bps); + } else { + // We give smoothed bitrate allocation to audio network adaptor as + // the uplink bandwidth. + // The BWE spikes should not affect the bitrate smoother more than 25%. + // To simplify the calculations we use a step response as input signal. + // The step response of an exponential filter is + // u(t) = 1 - e^(-t / time_constant). + // In order to limit the affect of a BWE spike within 25% of its value + // before + // the next BWE update, we would choose a time constant that fulfills + // 1 - e^(-bwe_period_ms / time_constant) < 0.25 + // Then 4 * bwe_period_ms is a good choice. + if (bwe_period_ms) + bitrate_smoother_->SetTimeConstantMs(*bwe_period_ms * 4); + bitrate_smoother_->AddSample(target_audio_bitrate_bps); + } + + ApplyAudioNetworkAdaptor(); + } else { + if (!overhead_bytes_per_packet_) { + RTC_LOG(LS_INFO) + << "AudioEncoderOpusImpl: Overhead unknown, target audio bitrate " + << target_audio_bitrate_bps << " bps is ignored."; + return; + } + const int overhead_bps = static_cast( + *overhead_bytes_per_packet_ * 8 * 100 / Num10MsFramesInNextPacket()); + SetTargetBitrate( + std::min(AudioEncoderOpusConfig::kMaxBitrateBps, + std::max(AudioEncoderOpusConfig::kMinBitrateBps, + target_audio_bitrate_bps - overhead_bps))); + } +} +void AudioEncoderOpusImpl::OnReceivedUplinkBandwidth( + int target_audio_bitrate_bps, + absl::optional bwe_period_ms) { + OnReceivedUplinkBandwidth(target_audio_bitrate_bps, bwe_period_ms, + absl::nullopt); +} + +void AudioEncoderOpusImpl::OnReceivedUplinkAllocation( + BitrateAllocationUpdate update) { + OnReceivedUplinkBandwidth(update.target_bitrate.bps(), update.bwe_period.ms(), + update.stable_target_bitrate.bps()); +} + +void AudioEncoderOpusImpl::OnReceivedRtt(int rtt_ms) { + if (!audio_network_adaptor_) + return; + audio_network_adaptor_->SetRtt(rtt_ms); + ApplyAudioNetworkAdaptor(); +} + +void AudioEncoderOpusImpl::OnReceivedOverhead( + size_t overhead_bytes_per_packet) { + if (audio_network_adaptor_) { + audio_network_adaptor_->SetOverhead(overhead_bytes_per_packet); + ApplyAudioNetworkAdaptor(); + } else { + overhead_bytes_per_packet_ = overhead_bytes_per_packet; + } +} + +void AudioEncoderOpusImpl::SetReceiverFrameLengthRange( + int min_frame_length_ms, + int max_frame_length_ms) { + // Ensure that `SetReceiverFrameLengthRange` is called before + // `EnableAudioNetworkAdaptor`, otherwise we need to recreate + // `audio_network_adaptor_`, which is not a needed use case. + RTC_DCHECK(!audio_network_adaptor_); + FindSupportedFrameLengths(min_frame_length_ms, max_frame_length_ms, + &config_.supported_frame_lengths_ms); +} + +AudioEncoder::EncodedInfo AudioEncoderOpusImpl::EncodeImpl( + uint32_t rtp_timestamp, + rtc::ArrayView audio, + rtc::Buffer* encoded) { + MaybeUpdateUplinkBandwidth(); + + if (input_buffer_.empty()) + first_timestamp_in_buffer_ = rtp_timestamp; + + input_buffer_.insert(input_buffer_.end(), audio.cbegin(), audio.cend()); + if (input_buffer_.size() < + (Num10msFramesPerPacket() * SamplesPer10msFrame())) { + return EncodedInfo(); + } + RTC_CHECK_EQ(input_buffer_.size(), + Num10msFramesPerPacket() * SamplesPer10msFrame()); + + const size_t max_encoded_bytes = SufficientOutputBufferSize(); + EncodedInfo info; + info.encoded_bytes = encoded->AppendData( + max_encoded_bytes, [&](rtc::ArrayView encoded) { + int status = WebRtcOpus_Encode( + inst_, &input_buffer_[0], + rtc::CheckedDivExact(input_buffer_.size(), config_.num_channels), + rtc::saturated_cast(max_encoded_bytes), encoded.data()); + + RTC_CHECK_GE(status, 0); // Fails only if fed invalid data. + + return static_cast(status); + }); + input_buffer_.clear(); + + bool dtx_frame = (info.encoded_bytes <= 2); + + // Will use new packet size for next encoding. + config_.frame_size_ms = next_frame_length_ms_; + + if (adjust_bandwidth_ && bitrate_changed_) { + const auto bandwidth = GetNewBandwidth(config_, inst_); + if (bandwidth) { + RTC_CHECK_EQ(0, WebRtcOpus_SetBandwidth(inst_, *bandwidth)); + } + bitrate_changed_ = false; + } + + info.encoded_timestamp = first_timestamp_in_buffer_; + info.payload_type = payload_type_; + info.send_even_if_empty = true; // Allows Opus to send empty packets. + // After 20 DTX frames (MAX_CONSECUTIVE_DTX) Opus will send a frame + // coding the background noise. Avoid flagging this frame as speech + // (even though there is a probability of the frame being speech). + info.speech = !dtx_frame && (consecutive_dtx_frames_ != 20); + info.encoder_type = CodecType::kOpus; + + // Increase or reset DTX counter. + consecutive_dtx_frames_ = (dtx_frame) ? (consecutive_dtx_frames_ + 1) : (0); + + return info; +} + +size_t AudioEncoderOpusImpl::Num10msFramesPerPacket() const { + return static_cast(rtc::CheckedDivExact(config_.frame_size_ms, 10)); +} + +size_t AudioEncoderOpusImpl::SamplesPer10msFrame() const { + return rtc::CheckedDivExact(config_.sample_rate_hz, 100) * + config_.num_channels; +} + +size_t AudioEncoderOpusImpl::SufficientOutputBufferSize() const { + // Calculate the number of bytes we expect the encoder to produce, + // then multiply by two to give a wide margin for error. + const size_t bytes_per_millisecond = + static_cast(GetBitrateBps(config_) / (1000 * 8) + 1); + const size_t approx_encoded_bytes = + Num10msFramesPerPacket() * 10 * bytes_per_millisecond; + return 2 * approx_encoded_bytes; +} + +// If the given config is OK, recreate the Opus encoder instance with those +// settings, save the config, and return true. Otherwise, do nothing and return +// false. +bool AudioEncoderOpusImpl::RecreateEncoderInstance( + const AudioEncoderOpusConfig& config) { + if (!config.IsOk()) + return false; + config_ = config; + if (inst_) + RTC_CHECK_EQ(0, WebRtcOpus_EncoderFree(inst_)); + input_buffer_.clear(); + input_buffer_.reserve(Num10msFramesPerPacket() * SamplesPer10msFrame()); + RTC_CHECK_EQ(0, WebRtcOpus_EncoderCreate( + &inst_, config.num_channels, + config.application == + AudioEncoderOpusConfig::ApplicationMode::kVoip + ? 0 + : 1, + config.sample_rate_hz)); + const int bitrate = GetBitrateBps(config); + RTC_CHECK_EQ(0, WebRtcOpus_SetBitRate(inst_, bitrate)); + RTC_LOG(LS_VERBOSE) << "Set Opus bitrate to " << bitrate << " bps."; + if (config.fec_enabled) { + RTC_CHECK_EQ(0, WebRtcOpus_EnableFec(inst_)); + } else { + RTC_CHECK_EQ(0, WebRtcOpus_DisableFec(inst_)); + } + RTC_CHECK_EQ( + 0, WebRtcOpus_SetMaxPlaybackRate(inst_, config.max_playback_rate_hz)); + // Use the default complexity if the start bitrate is within the hysteresis + // window. + complexity_ = GetNewComplexity(config).value_or(config.complexity); + RTC_CHECK_EQ(0, WebRtcOpus_SetComplexity(inst_, complexity_)); + bitrate_changed_ = true; + if (config.dtx_enabled) { + RTC_CHECK_EQ(0, WebRtcOpus_EnableDtx(inst_)); + } else { + RTC_CHECK_EQ(0, WebRtcOpus_DisableDtx(inst_)); + } + RTC_CHECK_EQ(0, + WebRtcOpus_SetPacketLossRate( + inst_, static_cast(packet_loss_rate_ * 100 + .5))); + if (config.cbr_enabled) { + RTC_CHECK_EQ(0, WebRtcOpus_EnableCbr(inst_)); + } else { + RTC_CHECK_EQ(0, WebRtcOpus_DisableCbr(inst_)); + } + num_channels_to_encode_ = NumChannels(); + next_frame_length_ms_ = config_.frame_size_ms; + return true; +} + +void AudioEncoderOpusImpl::SetFrameLength(int frame_length_ms) { + if (next_frame_length_ms_ != frame_length_ms) { + RTC_LOG(LS_VERBOSE) << "Update Opus frame length " + << "from " << next_frame_length_ms_ << " ms " + << "to " << frame_length_ms << " ms."; + } + next_frame_length_ms_ = frame_length_ms; +} + +void AudioEncoderOpusImpl::SetNumChannelsToEncode( + size_t num_channels_to_encode) { + RTC_DCHECK_GT(num_channels_to_encode, 0); + RTC_DCHECK_LE(num_channels_to_encode, config_.num_channels); + + if (num_channels_to_encode_ == num_channels_to_encode) + return; + + RTC_CHECK_EQ(0, WebRtcOpus_SetForceChannels(inst_, num_channels_to_encode)); + num_channels_to_encode_ = num_channels_to_encode; +} + +void AudioEncoderOpusImpl::SetProjectedPacketLossRate(float fraction) { + fraction = std::min(std::max(fraction, 0.0f), kMaxPacketLossFraction); + if (packet_loss_rate_ != fraction) { + packet_loss_rate_ = fraction; + RTC_CHECK_EQ( + 0, WebRtcOpus_SetPacketLossRate( + inst_, static_cast(packet_loss_rate_ * 100 + .5))); + } +} + +void AudioEncoderOpusImpl::SetTargetBitrate(int bits_per_second) { + const int new_bitrate = rtc::SafeClamp( + bits_per_second, AudioEncoderOpusConfig::kMinBitrateBps, + AudioEncoderOpusConfig::kMaxBitrateBps); + if (config_.bitrate_bps && *config_.bitrate_bps != new_bitrate) { + config_.bitrate_bps = new_bitrate; + RTC_DCHECK(config_.IsOk()); + const int bitrate = GetBitrateBps(config_); + RTC_CHECK_EQ( + 0, WebRtcOpus_SetBitRate( + inst_, GetMultipliedBitrate(bitrate, bitrate_multipliers_))); + RTC_LOG(LS_VERBOSE) << "Set Opus bitrate to " << bitrate << " bps."; + bitrate_changed_ = true; + } + + const auto new_complexity = GetNewComplexity(config_); + if (new_complexity && complexity_ != *new_complexity) { + complexity_ = *new_complexity; + RTC_CHECK_EQ(0, WebRtcOpus_SetComplexity(inst_, complexity_)); + } +} + +void AudioEncoderOpusImpl::ApplyAudioNetworkAdaptor() { + auto config = audio_network_adaptor_->GetEncoderRuntimeConfig(); + + if (config.bitrate_bps) + SetTargetBitrate(*config.bitrate_bps); + if (config.frame_length_ms) + SetFrameLength(*config.frame_length_ms); + if (config.enable_dtx) + SetDtx(*config.enable_dtx); + if (config.num_channels) + SetNumChannelsToEncode(*config.num_channels); +} + +std::unique_ptr +AudioEncoderOpusImpl::DefaultAudioNetworkAdaptorCreator( + absl::string_view config_string, + RtcEventLog* event_log) const { + AudioNetworkAdaptorImpl::Config config; + config.event_log = event_log; + return std::unique_ptr(new AudioNetworkAdaptorImpl( + config, ControllerManagerImpl::Create( + config_string, NumChannels(), supported_frame_lengths_ms(), + AudioEncoderOpusConfig::kMinBitrateBps, + num_channels_to_encode_, next_frame_length_ms_, + GetTargetBitrate(), config_.fec_enabled, GetDtx()))); +} + +void AudioEncoderOpusImpl::MaybeUpdateUplinkBandwidth() { + if (audio_network_adaptor_ && !use_stable_target_for_adaptation_) { + int64_t now_ms = rtc::TimeMillis(); + if (!bitrate_smoother_last_update_time_ || + now_ms - *bitrate_smoother_last_update_time_ >= + config_.uplink_bandwidth_update_interval_ms) { + absl::optional smoothed_bitrate = bitrate_smoother_->GetAverage(); + if (smoothed_bitrate) + audio_network_adaptor_->SetUplinkBandwidth(*smoothed_bitrate); + bitrate_smoother_last_update_time_ = now_ms; + } + } +} + +ANAStats AudioEncoderOpusImpl::GetANAStats() const { + if (audio_network_adaptor_) { + return audio_network_adaptor_->GetStats(); + } + return ANAStats(); +} + +absl::optional > +AudioEncoderOpusImpl::GetFrameLengthRange() const { + if (audio_network_adaptor_) { + if (config_.supported_frame_lengths_ms.empty()) { + return absl::nullopt; + } + return {{TimeDelta::Millis(config_.supported_frame_lengths_ms.front()), + TimeDelta::Millis(config_.supported_frame_lengths_ms.back())}}; + } else { + return {{TimeDelta::Millis(config_.frame_size_ms), + TimeDelta::Millis(config_.frame_size_ms)}}; + } +} + +} // namespace webrtc diff --git a/third_party/libwebrtc/modules/audio_coding/codecs/opus/audio_encoder_opus.h b/third_party/libwebrtc/modules/audio_coding/codecs/opus/audio_encoder_opus.h new file mode 100644 index 0000000000..8c5c235016 --- /dev/null +++ b/third_party/libwebrtc/modules/audio_coding/codecs/opus/audio_encoder_opus.h @@ -0,0 +1,184 @@ +/* + * Copyright (c) 2014 The WebRTC project authors. All Rights Reserved. + * + * Use of this source code is governed by a BSD-style license + * that can be found in the LICENSE file in the root of the source + * tree. An additional intellectual property rights grant can be found + * in the file PATENTS. All contributing project authors may + * be found in the AUTHORS file in the root of the source tree. + */ + +#ifndef MODULES_AUDIO_CODING_CODECS_OPUS_AUDIO_ENCODER_OPUS_H_ +#define MODULES_AUDIO_CODING_CODECS_OPUS_AUDIO_ENCODER_OPUS_H_ + +#include +#include +#include +#include + +#include "absl/strings/string_view.h" +#include "absl/types/optional.h" +#include "api/audio_codecs/audio_encoder.h" +#include "api/audio_codecs/audio_format.h" +#include "api/audio_codecs/opus/audio_encoder_opus_config.h" +#include "common_audio/smoothing_filter.h" +#include "modules/audio_coding/audio_network_adaptor/include/audio_network_adaptor.h" +#include "modules/audio_coding/codecs/opus/opus_interface.h" + +namespace webrtc { + +class RtcEventLog; + +class AudioEncoderOpusImpl final : public AudioEncoder { + public: + // Returns empty if the current bitrate falls within the hysteresis window, + // defined by complexity_threshold_bps +/- complexity_threshold_window_bps. + // Otherwise, returns the current complexity depending on whether the + // current bitrate is above or below complexity_threshold_bps. + static absl::optional GetNewComplexity( + const AudioEncoderOpusConfig& config); + + // Returns OPUS_AUTO if the the current bitrate is above wideband threshold. + // Returns empty if it is below, but bandwidth coincides with the desired one. + // Otherwise returns the desired bandwidth. + static absl::optional GetNewBandwidth( + const AudioEncoderOpusConfig& config, + OpusEncInst* inst); + + using AudioNetworkAdaptorCreator = + std::function(absl::string_view, + RtcEventLog*)>; + + AudioEncoderOpusImpl(const AudioEncoderOpusConfig& config, int payload_type); + + // Dependency injection for testing. + AudioEncoderOpusImpl( + const AudioEncoderOpusConfig& config, + int payload_type, + const AudioNetworkAdaptorCreator& audio_network_adaptor_creator, + std::unique_ptr bitrate_smoother); + + AudioEncoderOpusImpl(int payload_type, const SdpAudioFormat& format); + ~AudioEncoderOpusImpl() override; + + AudioEncoderOpusImpl(const AudioEncoderOpusImpl&) = delete; + AudioEncoderOpusImpl& operator=(const AudioEncoderOpusImpl&) = delete; + + int SampleRateHz() const override; + size_t NumChannels() const override; + int RtpTimestampRateHz() const override; + size_t Num10MsFramesInNextPacket() const override; + size_t Max10MsFramesInAPacket() const override; + int GetTargetBitrate() const override; + + void Reset() override; + bool SetFec(bool enable) override; + + // Set Opus DTX. Once enabled, Opus stops transmission, when it detects + // voice being inactive. During that, it still sends 2 packets (one for + // content, one for signaling) about every 400 ms. + bool SetDtx(bool enable) override; + bool GetDtx() const override; + + bool SetApplication(Application application) override; + void SetMaxPlaybackRate(int frequency_hz) override; + bool EnableAudioNetworkAdaptor(const std::string& config_string, + RtcEventLog* event_log) override; + void DisableAudioNetworkAdaptor() override; + void OnReceivedUplinkPacketLossFraction( + float uplink_packet_loss_fraction) override; + void OnReceivedTargetAudioBitrate(int target_audio_bitrate_bps) override; + void OnReceivedUplinkBandwidth( + int target_audio_bitrate_bps, + absl::optional bwe_period_ms) override; + void OnReceivedUplinkAllocation(BitrateAllocationUpdate update) override; + void OnReceivedRtt(int rtt_ms) override; + void OnReceivedOverhead(size_t overhead_bytes_per_packet) override; + void SetReceiverFrameLengthRange(int min_frame_length_ms, + int max_frame_length_ms) override; + ANAStats GetANAStats() const override; + absl::optional > GetFrameLengthRange() + const override; + rtc::ArrayView supported_frame_lengths_ms() const { + return config_.supported_frame_lengths_ms; + } + + // Getters for testing. + float packet_loss_rate() const { return packet_loss_rate_; } + AudioEncoderOpusConfig::ApplicationMode application() const { + return config_.application; + } + bool fec_enabled() const { return config_.fec_enabled; } + size_t num_channels_to_encode() const { return num_channels_to_encode_; } + int next_frame_length_ms() const { return next_frame_length_ms_; } + + protected: + EncodedInfo EncodeImpl(uint32_t rtp_timestamp, + rtc::ArrayView audio, + rtc::Buffer* encoded) override; + + private: + class PacketLossFractionSmoother; + + static absl::optional SdpToConfig( + const SdpAudioFormat& format); + static void AppendSupportedEncoders(std::vector* specs); + static AudioCodecInfo QueryAudioEncoder(const AudioEncoderOpusConfig& config); + static std::unique_ptr MakeAudioEncoder( + const AudioEncoderOpusConfig&, + int payload_type); + + size_t Num10msFramesPerPacket() const; + size_t SamplesPer10msFrame() const; + size_t SufficientOutputBufferSize() const; + bool RecreateEncoderInstance(const AudioEncoderOpusConfig& config); + void SetFrameLength(int frame_length_ms); + void SetNumChannelsToEncode(size_t num_channels_to_encode); + void SetProjectedPacketLossRate(float fraction); + + void OnReceivedUplinkBandwidth( + int target_audio_bitrate_bps, + absl::optional bwe_period_ms, + absl::optional link_capacity_allocation); + + // TODO(minyue): remove "override" when we can deprecate + // `AudioEncoder::SetTargetBitrate`. + void SetTargetBitrate(int target_bps) override; + + void ApplyAudioNetworkAdaptor(); + std::unique_ptr DefaultAudioNetworkAdaptorCreator( + absl::string_view config_string, + RtcEventLog* event_log) const; + + void MaybeUpdateUplinkBandwidth(); + + AudioEncoderOpusConfig config_; + const int payload_type_; + const bool use_stable_target_for_adaptation_; + const bool adjust_bandwidth_; + bool bitrate_changed_; + // A multiplier for bitrates at 5 kbps and higher. The target bitrate + // will be multiplied by these multipliers, each multiplier is applied to a + // 1 kbps range. + std::vector bitrate_multipliers_; + float packet_loss_rate_; + std::vector input_buffer_; + OpusEncInst* inst_; + uint32_t first_timestamp_in_buffer_; + size_t num_channels_to_encode_; + int next_frame_length_ms_; + int complexity_; + std::unique_ptr packet_loss_fraction_smoother_; + const AudioNetworkAdaptorCreator audio_network_adaptor_creator_; + std::unique_ptr audio_network_adaptor_; + absl::optional overhead_bytes_per_packet_; + const std::unique_ptr bitrate_smoother_; + absl::optional bitrate_smoother_last_update_time_; + int consecutive_dtx_frames_; + + friend struct AudioEncoderOpus; +}; + +} // namespace webrtc + +#endif // MODULES_AUDIO_CODING_CODECS_OPUS_AUDIO_ENCODER_OPUS_H_ diff --git a/third_party/libwebrtc/modules/audio_coding/codecs/opus/audio_encoder_opus_unittest.cc b/third_party/libwebrtc/modules/audio_coding/codecs/opus/audio_encoder_opus_unittest.cc new file mode 100644 index 0000000000..a2ebe43bbe --- /dev/null +++ b/third_party/libwebrtc/modules/audio_coding/codecs/opus/audio_encoder_opus_unittest.cc @@ -0,0 +1,914 @@ +/* + * Copyright (c) 2015 The WebRTC project authors. All Rights Reserved. + * + * Use of this source code is governed by a BSD-style license + * that can be found in the LICENSE file in the root of the source + * tree. An additional intellectual property rights grant can be found + * in the file PATENTS. All contributing project authors may + * be found in the AUTHORS file in the root of the source tree. + */ + +#include "api/audio_codecs/opus/audio_encoder_opus.h" + +#include +#include +#include + +#include "absl/strings/string_view.h" +#include "common_audio/mocks/mock_smoothing_filter.h" +#include "modules/audio_coding/audio_network_adaptor/mock/mock_audio_network_adaptor.h" +#include "modules/audio_coding/codecs/opus/audio_encoder_opus.h" +#include "modules/audio_coding/codecs/opus/opus_interface.h" +#include "modules/audio_coding/neteq/tools/audio_loop.h" +#include "rtc_base/checks.h" +#include "rtc_base/fake_clock.h" +#include "test/field_trial.h" +#include "test/gmock.h" +#include "test/gtest.h" +#include "test/testsupport/file_utils.h" + +namespace webrtc { +using ::testing::NiceMock; +using ::testing::Return; + +namespace { + +constexpr int kDefaultOpusPayloadType = 105; +constexpr int kDefaultOpusRate = 32000; +constexpr int kDefaultOpusPacSize = 960; +constexpr int64_t kInitialTimeUs = 12345678; + +AudioEncoderOpusConfig CreateConfigWithParameters( + const SdpAudioFormat::Parameters& params) { + const SdpAudioFormat format("opus", 48000, 2, params); + return *AudioEncoderOpus::SdpToConfig(format); +} + +struct AudioEncoderOpusStates { + MockAudioNetworkAdaptor* mock_audio_network_adaptor; + MockSmoothingFilter* mock_bitrate_smoother; + std::unique_ptr encoder; + std::unique_ptr fake_clock; + AudioEncoderOpusConfig config; +}; + +std::unique_ptr CreateCodec(int sample_rate_hz, + size_t num_channels) { + std::unique_ptr states = + std::make_unique(); + states->mock_audio_network_adaptor = nullptr; + states->fake_clock.reset(new rtc::ScopedFakeClock()); + states->fake_clock->SetTime(Timestamp::Micros(kInitialTimeUs)); + + MockAudioNetworkAdaptor** mock_ptr = &states->mock_audio_network_adaptor; + AudioEncoderOpusImpl::AudioNetworkAdaptorCreator creator = + [mock_ptr](absl::string_view, RtcEventLog* event_log) { + std::unique_ptr adaptor( + new NiceMock()); + EXPECT_CALL(*adaptor, Die()); + *mock_ptr = adaptor.get(); + return adaptor; + }; + + AudioEncoderOpusConfig config; + config.frame_size_ms = rtc::CheckedDivExact(kDefaultOpusPacSize, 48); + config.sample_rate_hz = sample_rate_hz; + config.num_channels = num_channels; + config.bitrate_bps = kDefaultOpusRate; + config.application = num_channels == 1 + ? AudioEncoderOpusConfig::ApplicationMode::kVoip + : AudioEncoderOpusConfig::ApplicationMode::kAudio; + config.supported_frame_lengths_ms.push_back(config.frame_size_ms); + states->config = config; + + std::unique_ptr bitrate_smoother( + new MockSmoothingFilter()); + states->mock_bitrate_smoother = bitrate_smoother.get(); + + states->encoder.reset( + new AudioEncoderOpusImpl(states->config, kDefaultOpusPayloadType, creator, + std::move(bitrate_smoother))); + return states; +} + +AudioEncoderRuntimeConfig CreateEncoderRuntimeConfig() { + constexpr int kBitrate = 40000; + constexpr int kFrameLength = 60; + constexpr bool kEnableDtx = false; + constexpr size_t kNumChannels = 1; + AudioEncoderRuntimeConfig config; + config.bitrate_bps = kBitrate; + config.frame_length_ms = kFrameLength; + config.enable_dtx = kEnableDtx; + config.num_channels = kNumChannels; + return config; +} + +void CheckEncoderRuntimeConfig(const AudioEncoderOpusImpl* encoder, + const AudioEncoderRuntimeConfig& config) { + EXPECT_EQ(*config.bitrate_bps, encoder->GetTargetBitrate()); + EXPECT_EQ(*config.frame_length_ms, encoder->next_frame_length_ms()); + EXPECT_EQ(*config.enable_dtx, encoder->GetDtx()); + EXPECT_EQ(*config.num_channels, encoder->num_channels_to_encode()); +} + +// Create 10ms audio data blocks for a total packet size of "packet_size_ms". +std::unique_ptr Create10msAudioBlocks( + const std::unique_ptr& encoder, + int packet_size_ms) { + const std::string file_name = + test::ResourcePath("audio_coding/testfile32kHz", "pcm"); + + std::unique_ptr speech_data(new test::AudioLoop()); + int audio_samples_per_ms = + rtc::CheckedDivExact(encoder->SampleRateHz(), 1000); + if (!speech_data->Init( + file_name, + packet_size_ms * audio_samples_per_ms * + encoder->num_channels_to_encode(), + 10 * audio_samples_per_ms * encoder->num_channels_to_encode())) + return nullptr; + return speech_data; +} + +} // namespace + +class AudioEncoderOpusTest : public ::testing::TestWithParam { + protected: + int sample_rate_hz_{GetParam()}; +}; +INSTANTIATE_TEST_SUITE_P(Param, + AudioEncoderOpusTest, + ::testing::Values(16000, 48000)); + +TEST_P(AudioEncoderOpusTest, DefaultApplicationModeMono) { + auto states = CreateCodec(sample_rate_hz_, 1); + EXPECT_EQ(AudioEncoderOpusConfig::ApplicationMode::kVoip, + states->encoder->application()); +} + +TEST_P(AudioEncoderOpusTest, DefaultApplicationModeStereo) { + auto states = CreateCodec(sample_rate_hz_, 2); + EXPECT_EQ(AudioEncoderOpusConfig::ApplicationMode::kAudio, + states->encoder->application()); +} + +TEST_P(AudioEncoderOpusTest, ChangeApplicationMode) { + auto states = CreateCodec(sample_rate_hz_, 2); + EXPECT_TRUE( + states->encoder->SetApplication(AudioEncoder::Application::kSpeech)); + EXPECT_EQ(AudioEncoderOpusConfig::ApplicationMode::kVoip, + states->encoder->application()); +} + +TEST_P(AudioEncoderOpusTest, ResetWontChangeApplicationMode) { + auto states = CreateCodec(sample_rate_hz_, 2); + + // Trigger a reset. + states->encoder->Reset(); + // Verify that the mode is still kAudio. + EXPECT_EQ(AudioEncoderOpusConfig::ApplicationMode::kAudio, + states->encoder->application()); + + // Now change to kVoip. + EXPECT_TRUE( + states->encoder->SetApplication(AudioEncoder::Application::kSpeech)); + EXPECT_EQ(AudioEncoderOpusConfig::ApplicationMode::kVoip, + states->encoder->application()); + + // Trigger a reset again. + states->encoder->Reset(); + // Verify that the mode is still kVoip. + EXPECT_EQ(AudioEncoderOpusConfig::ApplicationMode::kVoip, + states->encoder->application()); +} + +TEST_P(AudioEncoderOpusTest, ToggleDtx) { + auto states = CreateCodec(sample_rate_hz_, 2); + // Enable DTX + EXPECT_TRUE(states->encoder->SetDtx(true)); + EXPECT_TRUE(states->encoder->GetDtx()); + // Turn off DTX. + EXPECT_TRUE(states->encoder->SetDtx(false)); + EXPECT_FALSE(states->encoder->GetDtx()); +} + +TEST_P(AudioEncoderOpusTest, + OnReceivedUplinkBandwidthWithoutAudioNetworkAdaptor) { + auto states = CreateCodec(sample_rate_hz_, 1); + // Constants are replicated from audio_states->encoderopus.cc. + const int kMinBitrateBps = 6000; + const int kMaxBitrateBps = 510000; + const int kOverheadBytesPerPacket = 64; + states->encoder->OnReceivedOverhead(kOverheadBytesPerPacket); + const int kOverheadBps = 8 * kOverheadBytesPerPacket * + rtc::CheckedDivExact(48000, kDefaultOpusPacSize); + // Set a too low bitrate. + states->encoder->OnReceivedUplinkBandwidth(kMinBitrateBps + kOverheadBps - 1, + absl::nullopt); + EXPECT_EQ(kMinBitrateBps, states->encoder->GetTargetBitrate()); + // Set a too high bitrate. + states->encoder->OnReceivedUplinkBandwidth(kMaxBitrateBps + kOverheadBps + 1, + absl::nullopt); + EXPECT_EQ(kMaxBitrateBps, states->encoder->GetTargetBitrate()); + // Set the minimum rate. + states->encoder->OnReceivedUplinkBandwidth(kMinBitrateBps + kOverheadBps, + absl::nullopt); + EXPECT_EQ(kMinBitrateBps, states->encoder->GetTargetBitrate()); + // Set the maximum rate. + states->encoder->OnReceivedUplinkBandwidth(kMaxBitrateBps + kOverheadBps, + absl::nullopt); + EXPECT_EQ(kMaxBitrateBps, states->encoder->GetTargetBitrate()); + // Set rates from kMaxBitrateBps up to 32000 bps. + for (int rate = kMinBitrateBps + kOverheadBps; rate <= 32000 + kOverheadBps; + rate += 1000) { + states->encoder->OnReceivedUplinkBandwidth(rate, absl::nullopt); + EXPECT_EQ(rate - kOverheadBps, states->encoder->GetTargetBitrate()); + } +} + +TEST_P(AudioEncoderOpusTest, SetReceiverFrameLengthRange) { + auto states = CreateCodec(sample_rate_hz_, 2); + // Before calling to `SetReceiverFrameLengthRange`, + // `supported_frame_lengths_ms` should contain only the frame length being + // used. + using ::testing::ElementsAre; + EXPECT_THAT(states->encoder->supported_frame_lengths_ms(), + ElementsAre(states->encoder->next_frame_length_ms())); + states->encoder->SetReceiverFrameLengthRange(0, 12345); + states->encoder->SetReceiverFrameLengthRange(21, 60); + EXPECT_THAT(states->encoder->supported_frame_lengths_ms(), + ElementsAre(40, 60)); + states->encoder->SetReceiverFrameLengthRange(20, 59); + EXPECT_THAT(states->encoder->supported_frame_lengths_ms(), + ElementsAre(20, 40)); +} + +TEST_P(AudioEncoderOpusTest, + InvokeAudioNetworkAdaptorOnReceivedUplinkPacketLossFraction) { + auto states = CreateCodec(sample_rate_hz_, 2); + states->encoder->EnableAudioNetworkAdaptor("", nullptr); + + auto config = CreateEncoderRuntimeConfig(); + EXPECT_CALL(*states->mock_audio_network_adaptor, GetEncoderRuntimeConfig()) + .WillOnce(Return(config)); + + // Since using mock audio network adaptor, any packet loss fraction is fine. + constexpr float kUplinkPacketLoss = 0.1f; + EXPECT_CALL(*states->mock_audio_network_adaptor, + SetUplinkPacketLossFraction(kUplinkPacketLoss)); + states->encoder->OnReceivedUplinkPacketLossFraction(kUplinkPacketLoss); + + CheckEncoderRuntimeConfig(states->encoder.get(), config); +} + +TEST_P(AudioEncoderOpusTest, + InvokeAudioNetworkAdaptorOnReceivedUplinkBandwidth) { + test::ScopedFieldTrials override_field_trials( + "WebRTC-Audio-StableTargetAdaptation/Disabled/"); + auto states = CreateCodec(sample_rate_hz_, 2); + states->encoder->EnableAudioNetworkAdaptor("", nullptr); + + auto config = CreateEncoderRuntimeConfig(); + EXPECT_CALL(*states->mock_audio_network_adaptor, GetEncoderRuntimeConfig()) + .WillOnce(Return(config)); + + // Since using mock audio network adaptor, any target audio bitrate is fine. + constexpr int kTargetAudioBitrate = 30000; + constexpr int64_t kProbingIntervalMs = 3000; + EXPECT_CALL(*states->mock_audio_network_adaptor, + SetTargetAudioBitrate(kTargetAudioBitrate)); + EXPECT_CALL(*states->mock_bitrate_smoother, + SetTimeConstantMs(kProbingIntervalMs * 4)); + EXPECT_CALL(*states->mock_bitrate_smoother, AddSample(kTargetAudioBitrate)); + states->encoder->OnReceivedUplinkBandwidth(kTargetAudioBitrate, + kProbingIntervalMs); + + CheckEncoderRuntimeConfig(states->encoder.get(), config); +} + +TEST_P(AudioEncoderOpusTest, + InvokeAudioNetworkAdaptorOnReceivedUplinkAllocation) { + auto states = CreateCodec(sample_rate_hz_, 2); + states->encoder->EnableAudioNetworkAdaptor("", nullptr); + + auto config = CreateEncoderRuntimeConfig(); + EXPECT_CALL(*states->mock_audio_network_adaptor, GetEncoderRuntimeConfig()) + .WillOnce(Return(config)); + + BitrateAllocationUpdate update; + update.target_bitrate = DataRate::BitsPerSec(30000); + update.stable_target_bitrate = DataRate::BitsPerSec(20000); + update.bwe_period = TimeDelta::Millis(200); + EXPECT_CALL(*states->mock_audio_network_adaptor, + SetTargetAudioBitrate(update.target_bitrate.bps())); + EXPECT_CALL(*states->mock_audio_network_adaptor, + SetUplinkBandwidth(update.stable_target_bitrate.bps())); + states->encoder->OnReceivedUplinkAllocation(update); + + CheckEncoderRuntimeConfig(states->encoder.get(), config); +} + +TEST_P(AudioEncoderOpusTest, InvokeAudioNetworkAdaptorOnReceivedRtt) { + auto states = CreateCodec(sample_rate_hz_, 2); + states->encoder->EnableAudioNetworkAdaptor("", nullptr); + + auto config = CreateEncoderRuntimeConfig(); + EXPECT_CALL(*states->mock_audio_network_adaptor, GetEncoderRuntimeConfig()) + .WillOnce(Return(config)); + + // Since using mock audio network adaptor, any rtt is fine. + constexpr int kRtt = 30; + EXPECT_CALL(*states->mock_audio_network_adaptor, SetRtt(kRtt)); + states->encoder->OnReceivedRtt(kRtt); + + CheckEncoderRuntimeConfig(states->encoder.get(), config); +} + +TEST_P(AudioEncoderOpusTest, InvokeAudioNetworkAdaptorOnReceivedOverhead) { + auto states = CreateCodec(sample_rate_hz_, 2); + states->encoder->EnableAudioNetworkAdaptor("", nullptr); + + auto config = CreateEncoderRuntimeConfig(); + EXPECT_CALL(*states->mock_audio_network_adaptor, GetEncoderRuntimeConfig()) + .WillOnce(Return(config)); + + // Since using mock audio network adaptor, any overhead is fine. + constexpr size_t kOverhead = 64; + EXPECT_CALL(*states->mock_audio_network_adaptor, SetOverhead(kOverhead)); + states->encoder->OnReceivedOverhead(kOverhead); + + CheckEncoderRuntimeConfig(states->encoder.get(), config); +} + +TEST_P(AudioEncoderOpusTest, + PacketLossFractionSmoothedOnSetUplinkPacketLossFraction) { + auto states = CreateCodec(sample_rate_hz_, 2); + + // The values are carefully chosen so that if no smoothing is made, the test + // will fail. + constexpr float kPacketLossFraction_1 = 0.02f; + constexpr float kPacketLossFraction_2 = 0.198f; + // `kSecondSampleTimeMs` is chosen to ease the calculation since + // 0.9999 ^ 6931 = 0.5. + constexpr int64_t kSecondSampleTimeMs = 6931; + + // First time, no filtering. + states->encoder->OnReceivedUplinkPacketLossFraction(kPacketLossFraction_1); + EXPECT_FLOAT_EQ(0.02f, states->encoder->packet_loss_rate()); + + states->fake_clock->AdvanceTime(TimeDelta::Millis(kSecondSampleTimeMs)); + states->encoder->OnReceivedUplinkPacketLossFraction(kPacketLossFraction_2); + + // Now the output of packet loss fraction smoother should be + // (0.02 + 0.198) / 2 = 0.109. + EXPECT_NEAR(0.109f, states->encoder->packet_loss_rate(), 0.001); +} + +TEST_P(AudioEncoderOpusTest, PacketLossRateUpperBounded) { + auto states = CreateCodec(sample_rate_hz_, 2); + + states->encoder->OnReceivedUplinkPacketLossFraction(0.5); + EXPECT_FLOAT_EQ(0.2f, states->encoder->packet_loss_rate()); +} + +TEST_P(AudioEncoderOpusTest, DoNotInvokeSetTargetBitrateIfOverheadUnknown) { + auto states = CreateCodec(sample_rate_hz_, 2); + + states->encoder->OnReceivedUplinkBandwidth(kDefaultOpusRate * 2, + absl::nullopt); + + // Since `OnReceivedOverhead` has not been called, the codec bitrate should + // not change. + EXPECT_EQ(kDefaultOpusRate, states->encoder->GetTargetBitrate()); +} + +// Verifies that the complexity adaptation in the config works as intended. +TEST(AudioEncoderOpusTest, ConfigComplexityAdaptation) { + AudioEncoderOpusConfig config; + config.low_rate_complexity = 8; + config.complexity = 6; + + // Bitrate within hysteresis window. Expect empty output. + config.bitrate_bps = 12500; + EXPECT_EQ(absl::nullopt, AudioEncoderOpusImpl::GetNewComplexity(config)); + + // Bitrate below hysteresis window. Expect higher complexity. + config.bitrate_bps = 10999; + EXPECT_EQ(8, AudioEncoderOpusImpl::GetNewComplexity(config)); + + // Bitrate within hysteresis window. Expect empty output. + config.bitrate_bps = 12500; + EXPECT_EQ(absl::nullopt, AudioEncoderOpusImpl::GetNewComplexity(config)); + + // Bitrate above hysteresis window. Expect lower complexity. + config.bitrate_bps = 14001; + EXPECT_EQ(6, AudioEncoderOpusImpl::GetNewComplexity(config)); +} + +// Verifies that the bandwidth adaptation in the config works as intended. +TEST_P(AudioEncoderOpusTest, ConfigBandwidthAdaptation) { + AudioEncoderOpusConfig config; + const size_t opus_rate_khz = rtc::CheckedDivExact(sample_rate_hz_, 1000); + const std::vector silence( + opus_rate_khz * config.frame_size_ms * config.num_channels, 0); + constexpr size_t kMaxBytes = 1000; + uint8_t bitstream[kMaxBytes]; + + OpusEncInst* inst; + EXPECT_EQ(0, WebRtcOpus_EncoderCreate( + &inst, config.num_channels, + config.application == + AudioEncoderOpusConfig::ApplicationMode::kVoip + ? 0 + : 1, + sample_rate_hz_)); + + // Bitrate below minmum wideband. Expect narrowband. + config.bitrate_bps = absl::optional(7999); + auto bandwidth = AudioEncoderOpusImpl::GetNewBandwidth(config, inst); + EXPECT_EQ(absl::optional(OPUS_BANDWIDTH_NARROWBAND), bandwidth); + WebRtcOpus_SetBandwidth(inst, *bandwidth); + // It is necessary to encode here because Opus has some logic in the encoder + // that goes from the user-set bandwidth to the used and returned one. + WebRtcOpus_Encode(inst, silence.data(), + rtc::CheckedDivExact(silence.size(), config.num_channels), + kMaxBytes, bitstream); + + // Bitrate not yet above maximum narrowband. Expect empty. + config.bitrate_bps = absl::optional(9000); + bandwidth = AudioEncoderOpusImpl::GetNewBandwidth(config, inst); + EXPECT_EQ(absl::optional(), bandwidth); + + // Bitrate above maximum narrowband. Expect wideband. + config.bitrate_bps = absl::optional(9001); + bandwidth = AudioEncoderOpusImpl::GetNewBandwidth(config, inst); + EXPECT_EQ(absl::optional(OPUS_BANDWIDTH_WIDEBAND), bandwidth); + WebRtcOpus_SetBandwidth(inst, *bandwidth); + // It is necessary to encode here because Opus has some logic in the encoder + // that goes from the user-set bandwidth to the used and returned one. + WebRtcOpus_Encode(inst, silence.data(), + rtc::CheckedDivExact(silence.size(), config.num_channels), + kMaxBytes, bitstream); + + // Bitrate not yet below minimum wideband. Expect empty. + config.bitrate_bps = absl::optional(8000); + bandwidth = AudioEncoderOpusImpl::GetNewBandwidth(config, inst); + EXPECT_EQ(absl::optional(), bandwidth); + + // Bitrate above automatic threshold. Expect automatic. + config.bitrate_bps = absl::optional(12001); + bandwidth = AudioEncoderOpusImpl::GetNewBandwidth(config, inst); + EXPECT_EQ(absl::optional(OPUS_AUTO), bandwidth); + + EXPECT_EQ(0, WebRtcOpus_EncoderFree(inst)); +} + +TEST_P(AudioEncoderOpusTest, EmptyConfigDoesNotAffectEncoderSettings) { + auto states = CreateCodec(sample_rate_hz_, 2); + states->encoder->EnableAudioNetworkAdaptor("", nullptr); + + auto config = CreateEncoderRuntimeConfig(); + AudioEncoderRuntimeConfig empty_config; + + EXPECT_CALL(*states->mock_audio_network_adaptor, GetEncoderRuntimeConfig()) + .WillOnce(Return(config)) + .WillOnce(Return(empty_config)); + + constexpr size_t kOverhead = 64; + EXPECT_CALL(*states->mock_audio_network_adaptor, SetOverhead(kOverhead)) + .Times(2); + states->encoder->OnReceivedOverhead(kOverhead); + states->encoder->OnReceivedOverhead(kOverhead); + + CheckEncoderRuntimeConfig(states->encoder.get(), config); +} + +TEST_P(AudioEncoderOpusTest, UpdateUplinkBandwidthInAudioNetworkAdaptor) { + test::ScopedFieldTrials override_field_trials( + "WebRTC-Audio-StableTargetAdaptation/Disabled/"); + auto states = CreateCodec(sample_rate_hz_, 2); + states->encoder->EnableAudioNetworkAdaptor("", nullptr); + const size_t opus_rate_khz = rtc::CheckedDivExact(sample_rate_hz_, 1000); + const std::vector audio(opus_rate_khz * 10 * 2, 0); + rtc::Buffer encoded; + EXPECT_CALL(*states->mock_bitrate_smoother, GetAverage()) + .WillOnce(Return(50000)); + EXPECT_CALL(*states->mock_audio_network_adaptor, SetUplinkBandwidth(50000)); + states->encoder->Encode( + 0, rtc::ArrayView(audio.data(), audio.size()), &encoded); + + // Repeat update uplink bandwidth tests. + for (int i = 0; i < 5; i++) { + // Don't update till it is time to update again. + states->fake_clock->AdvanceTime(TimeDelta::Millis( + states->config.uplink_bandwidth_update_interval_ms - 1)); + states->encoder->Encode( + 0, rtc::ArrayView(audio.data(), audio.size()), &encoded); + + // Update when it is time to update. + EXPECT_CALL(*states->mock_bitrate_smoother, GetAverage()) + .WillOnce(Return(40000)); + EXPECT_CALL(*states->mock_audio_network_adaptor, SetUplinkBandwidth(40000)); + states->fake_clock->AdvanceTime(TimeDelta::Millis(1)); + states->encoder->Encode( + 0, rtc::ArrayView(audio.data(), audio.size()), &encoded); + } +} + +TEST_P(AudioEncoderOpusTest, EncodeAtMinBitrate) { + auto states = CreateCodec(sample_rate_hz_, 1); + constexpr int kNumPacketsToEncode = 2; + auto audio_frames = + Create10msAudioBlocks(states->encoder, kNumPacketsToEncode * 20); + ASSERT_TRUE(audio_frames) << "Create10msAudioBlocks failed"; + rtc::Buffer encoded; + uint32_t rtp_timestamp = 12345; // Just a number not important to this test. + + states->encoder->OnReceivedUplinkBandwidth(0, absl::nullopt); + for (int packet_index = 0; packet_index < kNumPacketsToEncode; + packet_index++) { + // Make sure we are not encoding before we have enough data for + // a 20ms packet. + for (int index = 0; index < 1; index++) { + states->encoder->Encode(rtp_timestamp, audio_frames->GetNextBlock(), + &encoded); + EXPECT_EQ(0u, encoded.size()); + } + + // Should encode now. + states->encoder->Encode(rtp_timestamp, audio_frames->GetNextBlock(), + &encoded); + EXPECT_GT(encoded.size(), 0u); + encoded.Clear(); + } +} + +TEST(AudioEncoderOpusTest, TestConfigDefaults) { + const auto config_opt = AudioEncoderOpus::SdpToConfig({"opus", 48000, 2}); + ASSERT_TRUE(config_opt); + EXPECT_EQ(48000, config_opt->max_playback_rate_hz); + EXPECT_EQ(1u, config_opt->num_channels); + EXPECT_FALSE(config_opt->fec_enabled); + EXPECT_FALSE(config_opt->dtx_enabled); + EXPECT_EQ(20, config_opt->frame_size_ms); +} + +TEST(AudioEncoderOpusTest, TestConfigFromParams) { + const auto config1 = CreateConfigWithParameters({{"stereo", "0"}}); + EXPECT_EQ(1U, config1.num_channels); + + const auto config2 = CreateConfigWithParameters({{"stereo", "1"}}); + EXPECT_EQ(2U, config2.num_channels); + + const auto config3 = CreateConfigWithParameters({{"useinbandfec", "0"}}); + EXPECT_FALSE(config3.fec_enabled); + + const auto config4 = CreateConfigWithParameters({{"useinbandfec", "1"}}); + EXPECT_TRUE(config4.fec_enabled); + + const auto config5 = CreateConfigWithParameters({{"usedtx", "0"}}); + EXPECT_FALSE(config5.dtx_enabled); + + const auto config6 = CreateConfigWithParameters({{"usedtx", "1"}}); + EXPECT_TRUE(config6.dtx_enabled); + + const auto config7 = CreateConfigWithParameters({{"cbr", "0"}}); + EXPECT_FALSE(config7.cbr_enabled); + + const auto config8 = CreateConfigWithParameters({{"cbr", "1"}}); + EXPECT_TRUE(config8.cbr_enabled); + + const auto config9 = + CreateConfigWithParameters({{"maxplaybackrate", "12345"}}); + EXPECT_EQ(12345, config9.max_playback_rate_hz); + + const auto config10 = + CreateConfigWithParameters({{"maxaveragebitrate", "96000"}}); + EXPECT_EQ(96000, config10.bitrate_bps); + + const auto config11 = CreateConfigWithParameters({{"maxptime", "40"}}); + for (int frame_length : config11.supported_frame_lengths_ms) { + EXPECT_LE(frame_length, 40); + } + + const auto config12 = CreateConfigWithParameters({{"minptime", "40"}}); + for (int frame_length : config12.supported_frame_lengths_ms) { + EXPECT_GE(frame_length, 40); + } + + const auto config13 = CreateConfigWithParameters({{"ptime", "40"}}); + EXPECT_EQ(40, config13.frame_size_ms); + + constexpr int kMinSupportedFrameLength = 10; + constexpr int kMaxSupportedFrameLength = + WEBRTC_OPUS_SUPPORT_120MS_PTIME ? 120 : 60; + + const auto config14 = CreateConfigWithParameters({{"ptime", "1"}}); + EXPECT_EQ(kMinSupportedFrameLength, config14.frame_size_ms); + + const auto config15 = CreateConfigWithParameters({{"ptime", "2000"}}); + EXPECT_EQ(kMaxSupportedFrameLength, config15.frame_size_ms); +} + +TEST(AudioEncoderOpusTest, TestConfigFromInvalidParams) { + const webrtc::SdpAudioFormat format("opus", 48000, 2); + const auto default_config = *AudioEncoderOpus::SdpToConfig(format); +#if WEBRTC_OPUS_SUPPORT_120MS_PTIME + const std::vector default_supported_frame_lengths_ms({20, 40, 60, 120}); +#else + const std::vector default_supported_frame_lengths_ms({20, 40, 60}); +#endif + + AudioEncoderOpusConfig config; + config = CreateConfigWithParameters({{"stereo", "invalid"}}); + EXPECT_EQ(default_config.num_channels, config.num_channels); + + config = CreateConfigWithParameters({{"useinbandfec", "invalid"}}); + EXPECT_EQ(default_config.fec_enabled, config.fec_enabled); + + config = CreateConfigWithParameters({{"usedtx", "invalid"}}); + EXPECT_EQ(default_config.dtx_enabled, config.dtx_enabled); + + config = CreateConfigWithParameters({{"cbr", "invalid"}}); + EXPECT_EQ(default_config.dtx_enabled, config.dtx_enabled); + + config = CreateConfigWithParameters({{"maxplaybackrate", "0"}}); + EXPECT_EQ(default_config.max_playback_rate_hz, config.max_playback_rate_hz); + + config = CreateConfigWithParameters({{"maxplaybackrate", "-23"}}); + EXPECT_EQ(default_config.max_playback_rate_hz, config.max_playback_rate_hz); + + config = CreateConfigWithParameters({{"maxplaybackrate", "not a number!"}}); + EXPECT_EQ(default_config.max_playback_rate_hz, config.max_playback_rate_hz); + + config = CreateConfigWithParameters({{"maxaveragebitrate", "0"}}); + EXPECT_EQ(6000, config.bitrate_bps); + + config = CreateConfigWithParameters({{"maxaveragebitrate", "-1000"}}); + EXPECT_EQ(6000, config.bitrate_bps); + + config = CreateConfigWithParameters({{"maxaveragebitrate", "1024000"}}); + EXPECT_EQ(510000, config.bitrate_bps); + + config = CreateConfigWithParameters({{"maxaveragebitrate", "not a number!"}}); + EXPECT_EQ(default_config.bitrate_bps, config.bitrate_bps); + + config = CreateConfigWithParameters({{"maxptime", "invalid"}}); + EXPECT_EQ(default_supported_frame_lengths_ms, + config.supported_frame_lengths_ms); + + config = CreateConfigWithParameters({{"minptime", "invalid"}}); + EXPECT_EQ(default_supported_frame_lengths_ms, + config.supported_frame_lengths_ms); + + config = CreateConfigWithParameters({{"ptime", "invalid"}}); + EXPECT_EQ(default_supported_frame_lengths_ms, + config.supported_frame_lengths_ms); +} + +TEST(AudioEncoderOpusTest, GetFrameLenghtRange) { + AudioEncoderOpusConfig config = + CreateConfigWithParameters({{"maxptime", "10"}, {"ptime", "10"}}); + std::unique_ptr encoder = + AudioEncoderOpus::MakeAudioEncoder(config, kDefaultOpusPayloadType); + auto ptime = webrtc::TimeDelta::Millis(10); + absl::optional> range = { + {ptime, ptime}}; + EXPECT_EQ(encoder->GetFrameLengthRange(), range); +} + +// Test that bitrate will be overridden by the "maxaveragebitrate" parameter. +// Also test that the "maxaveragebitrate" can't be set to values outside the +// range of 6000 and 510000 +TEST(AudioEncoderOpusTest, SetSendCodecOpusMaxAverageBitrate) { + // Ignore if less than 6000. + const auto config1 = AudioEncoderOpus::SdpToConfig( + {"opus", 48000, 2, {{"maxaveragebitrate", "5999"}}}); + EXPECT_EQ(6000, config1->bitrate_bps); + + // Ignore if larger than 510000. + const auto config2 = AudioEncoderOpus::SdpToConfig( + {"opus", 48000, 2, {{"maxaveragebitrate", "510001"}}}); + EXPECT_EQ(510000, config2->bitrate_bps); + + const auto config3 = AudioEncoderOpus::SdpToConfig( + {"opus", 48000, 2, {{"maxaveragebitrate", "200000"}}}); + EXPECT_EQ(200000, config3->bitrate_bps); +} + +// Test maxplaybackrate <= 8000 triggers Opus narrow band mode. +TEST(AudioEncoderOpusTest, SetMaxPlaybackRateNb) { + auto config = CreateConfigWithParameters({{"maxplaybackrate", "8000"}}); + EXPECT_EQ(8000, config.max_playback_rate_hz); + EXPECT_EQ(12000, config.bitrate_bps); + + config = CreateConfigWithParameters( + {{"maxplaybackrate", "8000"}, {"stereo", "1"}}); + EXPECT_EQ(8000, config.max_playback_rate_hz); + EXPECT_EQ(24000, config.bitrate_bps); +} + +// Test 8000 < maxplaybackrate <= 12000 triggers Opus medium band mode. +TEST(AudioEncoderOpusTest, SetMaxPlaybackRateMb) { + auto config = CreateConfigWithParameters({{"maxplaybackrate", "8001"}}); + EXPECT_EQ(8001, config.max_playback_rate_hz); + EXPECT_EQ(20000, config.bitrate_bps); + + config = CreateConfigWithParameters( + {{"maxplaybackrate", "8001"}, {"stereo", "1"}}); + EXPECT_EQ(8001, config.max_playback_rate_hz); + EXPECT_EQ(40000, config.bitrate_bps); +} + +// Test 12000 < maxplaybackrate <= 16000 triggers Opus wide band mode. +TEST(AudioEncoderOpusTest, SetMaxPlaybackRateWb) { + auto config = CreateConfigWithParameters({{"maxplaybackrate", "12001"}}); + EXPECT_EQ(12001, config.max_playback_rate_hz); + EXPECT_EQ(20000, config.bitrate_bps); + + config = CreateConfigWithParameters( + {{"maxplaybackrate", "12001"}, {"stereo", "1"}}); + EXPECT_EQ(12001, config.max_playback_rate_hz); + EXPECT_EQ(40000, config.bitrate_bps); +} + +// Test 16000 < maxplaybackrate <= 24000 triggers Opus super wide band mode. +TEST(AudioEncoderOpusTest, SetMaxPlaybackRateSwb) { + auto config = CreateConfigWithParameters({{"maxplaybackrate", "16001"}}); + EXPECT_EQ(16001, config.max_playback_rate_hz); + EXPECT_EQ(32000, config.bitrate_bps); + + config = CreateConfigWithParameters( + {{"maxplaybackrate", "16001"}, {"stereo", "1"}}); + EXPECT_EQ(16001, config.max_playback_rate_hz); + EXPECT_EQ(64000, config.bitrate_bps); +} + +// Test 24000 < maxplaybackrate triggers Opus full band mode. +TEST(AudioEncoderOpusTest, SetMaxPlaybackRateFb) { + auto config = CreateConfigWithParameters({{"maxplaybackrate", "24001"}}); + EXPECT_EQ(24001, config.max_playback_rate_hz); + EXPECT_EQ(32000, config.bitrate_bps); + + config = CreateConfigWithParameters( + {{"maxplaybackrate", "24001"}, {"stereo", "1"}}); + EXPECT_EQ(24001, config.max_playback_rate_hz); + EXPECT_EQ(64000, config.bitrate_bps); +} + +TEST_P(AudioEncoderOpusTest, OpusFlagDtxAsNonSpeech) { + // Create encoder with DTX enabled. + AudioEncoderOpusConfig config; + config.dtx_enabled = true; + config.sample_rate_hz = sample_rate_hz_; + constexpr int payload_type = 17; + const auto encoder = AudioEncoderOpus::MakeAudioEncoder(config, payload_type); + + // Open file containing speech and silence. + const std::string kInputFileName = + webrtc::test::ResourcePath("audio_coding/testfile32kHz", "pcm"); + test::AudioLoop audio_loop; + // Use the file as if it were sampled at our desired input rate. + const size_t max_loop_length_samples = + sample_rate_hz_ * 10; // Max 10 second loop. + const size_t input_block_size_samples = + 10 * sample_rate_hz_ / 1000; // 10 ms. + EXPECT_TRUE(audio_loop.Init(kInputFileName, max_loop_length_samples, + input_block_size_samples)); + + // Encode. + AudioEncoder::EncodedInfo info; + rtc::Buffer encoded(500); + int nonspeech_frames = 0; + int max_nonspeech_frames = 0; + int dtx_frames = 0; + int max_dtx_frames = 0; + uint32_t rtp_timestamp = 0u; + for (size_t i = 0; i < 500; ++i) { + encoded.Clear(); + + // Every second call to the encoder will generate an Opus packet. + for (int j = 0; j < 2; j++) { + info = + encoder->Encode(rtp_timestamp, audio_loop.GetNextBlock(), &encoded); + rtp_timestamp += input_block_size_samples; + } + + // Bookkeeping of number of DTX frames. + if (info.encoded_bytes <= 2) { + ++dtx_frames; + } else { + if (dtx_frames > max_dtx_frames) + max_dtx_frames = dtx_frames; + dtx_frames = 0; + } + + // Bookkeeping of number of non-speech frames. + if (info.speech == 0) { + ++nonspeech_frames; + } else { + if (nonspeech_frames > max_nonspeech_frames) + max_nonspeech_frames = nonspeech_frames; + nonspeech_frames = 0; + } + } + + // Maximum number of consecutive non-speech packets should exceed 15. + EXPECT_GT(max_nonspeech_frames, 15); +} + +TEST(AudioEncoderOpusTest, OpusDtxFilteringHighEnergyRefreshPackets) { + test::ScopedFieldTrials override_field_trials( + "WebRTC-Audio-OpusAvoidNoisePumpingDuringDtx/Enabled/"); + const std::string kInputFileName = + webrtc::test::ResourcePath("audio_coding/testfile16kHz", "pcm"); + constexpr int kSampleRateHz = 16000; + AudioEncoderOpusConfig config; + config.dtx_enabled = true; + config.sample_rate_hz = kSampleRateHz; + constexpr int payload_type = 17; + const auto encoder = AudioEncoderOpus::MakeAudioEncoder(config, payload_type); + test::AudioLoop audio_loop; + constexpr size_t kMaxLoopLengthSaples = kSampleRateHz * 11.6f; + constexpr size_t kInputBlockSizeSamples = kSampleRateHz / 100; + EXPECT_TRUE(audio_loop.Init(kInputFileName, kMaxLoopLengthSaples, + kInputBlockSizeSamples)); + AudioEncoder::EncodedInfo info; + rtc::Buffer encoded(500); + // Encode the audio file and store the last part that corresponds to silence. + constexpr size_t kSilenceDurationSamples = kSampleRateHz * 0.2f; + std::array silence; + uint32_t rtp_timestamp = 0; + bool last_packet_dtx_frame = false; + bool opus_entered_dtx = false; + bool silence_filled = false; + size_t timestamp_start_silence = 0; + while (!silence_filled && rtp_timestamp < kMaxLoopLengthSaples) { + encoded.Clear(); + // Every second call to the encoder will generate an Opus packet. + for (int j = 0; j < 2; j++) { + auto next_frame = audio_loop.GetNextBlock(); + info = encoder->Encode(rtp_timestamp, next_frame, &encoded); + if (opus_entered_dtx) { + size_t silence_frame_start = rtp_timestamp - timestamp_start_silence; + silence_filled = silence_frame_start >= kSilenceDurationSamples; + if (!silence_filled) { + std::copy(next_frame.begin(), next_frame.end(), + silence.begin() + silence_frame_start); + } + } + rtp_timestamp += kInputBlockSizeSamples; + } + EXPECT_TRUE(info.encoded_bytes > 0 || last_packet_dtx_frame); + last_packet_dtx_frame = info.encoded_bytes > 0 ? info.encoded_bytes <= 2 + : last_packet_dtx_frame; + if (info.encoded_bytes <= 2 && !opus_entered_dtx) { + timestamp_start_silence = rtp_timestamp; + } + opus_entered_dtx = info.encoded_bytes <= 2; + } + + EXPECT_TRUE(silence_filled); + // The copied 200 ms of silence is used for creating 6 bursts that are fed to + // the encoder, the first three ones with a larger energy and the last three + // with a lower energy. This test verifies that the encoder just sends refresh + // DTX packets during the last bursts. + int number_non_empty_packets_during_increase = 0; + int number_non_empty_packets_during_decrease = 0; + for (size_t burst = 0; burst < 6; ++burst) { + uint32_t rtp_timestamp_start = rtp_timestamp; + const bool increase_noise = burst < 3; + const float gain = increase_noise ? 1.4f : 0.0f; + while (rtp_timestamp < rtp_timestamp_start + kSilenceDurationSamples) { + encoded.Clear(); + // Every second call to the encoder will generate an Opus packet. + for (int j = 0; j < 2; j++) { + std::array silence_frame; + size_t silence_frame_start = rtp_timestamp - rtp_timestamp_start; + std::transform( + silence.begin() + silence_frame_start, + silence.begin() + silence_frame_start + kInputBlockSizeSamples, + silence_frame.begin(), [gain](float s) { return gain * s; }); + info = encoder->Encode(rtp_timestamp, silence_frame, &encoded); + rtp_timestamp += kInputBlockSizeSamples; + } + EXPECT_TRUE(info.encoded_bytes > 0 || last_packet_dtx_frame); + last_packet_dtx_frame = info.encoded_bytes > 0 ? info.encoded_bytes <= 2 + : last_packet_dtx_frame; + // Tracking the number of non empty packets. + if (increase_noise && info.encoded_bytes > 2) { + number_non_empty_packets_during_increase++; + } + if (!increase_noise && info.encoded_bytes > 2) { + number_non_empty_packets_during_decrease++; + } + } + } + // Check that the refresh DTX packets are just sent during the decrease energy + // region. + EXPECT_EQ(number_non_empty_packets_during_increase, 0); + EXPECT_GT(number_non_empty_packets_during_decrease, 0); +} + +} // namespace webrtc diff --git a/third_party/libwebrtc/modules/audio_coding/codecs/opus/opus_bandwidth_unittest.cc b/third_party/libwebrtc/modules/audio_coding/codecs/opus/opus_bandwidth_unittest.cc new file mode 100644 index 0000000000..38b60c6187 --- /dev/null +++ b/third_party/libwebrtc/modules/audio_coding/codecs/opus/opus_bandwidth_unittest.cc @@ -0,0 +1,152 @@ +/* + * Copyright (c) 2017 The WebRTC project authors. All Rights Reserved. + * + * Use of this source code is governed by a BSD-style license + * that can be found in the LICENSE file in the root of the source + * tree. An additional intellectual property rights grant can be found + * in the file PATENTS. All contributing project authors may + * be found in the AUTHORS file in the root of the source tree. + */ + +#include "api/audio_codecs/opus/audio_decoder_opus.h" +#include "api/audio_codecs/opus/audio_encoder_opus.h" +#include "common_audio/include/audio_util.h" +#include "common_audio/window_generator.h" +#include "modules/audio_coding/codecs/opus/test/lapped_transform.h" +#include "modules/audio_coding/neteq/tools/audio_loop.h" +#include "test/field_trial.h" +#include "test/gtest.h" +#include "test/testsupport/file_utils.h" + +namespace webrtc { +namespace { + +constexpr size_t kNumChannels = 1u; +constexpr int kSampleRateHz = 48000; +constexpr size_t kMaxLoopLengthSamples = kSampleRateHz * 50; // 50 seconds. +constexpr size_t kInputBlockSizeSamples = 10 * kSampleRateHz / 1000; // 10 ms +constexpr size_t kOutputBlockSizeSamples = 20 * kSampleRateHz / 1000; // 20 ms +constexpr size_t kFftSize = 1024; +constexpr size_t kNarrowbandSize = 4000 * kFftSize / kSampleRateHz; +constexpr float kKbdAlpha = 1.5f; + +class PowerRatioEstimator : public LappedTransform::Callback { + public: + PowerRatioEstimator() : low_pow_(0.f), high_pow_(0.f) { + WindowGenerator::KaiserBesselDerived(kKbdAlpha, kFftSize, window_); + transform_.reset(new LappedTransform(kNumChannels, 0u, + kInputBlockSizeSamples, window_, + kFftSize, kFftSize / 2, this)); + } + + void ProcessBlock(float* data) { transform_->ProcessChunk(&data, nullptr); } + + float PowerRatio() { return high_pow_ / low_pow_; } + + protected: + void ProcessAudioBlock(const std::complex* const* input, + size_t num_input_channels, + size_t num_freq_bins, + size_t num_output_channels, + std::complex* const* output) override { + float low_pow = 0.f; + float high_pow = 0.f; + for (size_t i = 0u; i < num_input_channels; ++i) { + for (size_t j = 0u; j < kNarrowbandSize; ++j) { + float low_mag = std::abs(input[i][j]); + low_pow += low_mag * low_mag; + float high_mag = std::abs(input[i][j + kNarrowbandSize]); + high_pow += high_mag * high_mag; + } + } + low_pow_ += low_pow / (num_input_channels * kFftSize); + high_pow_ += high_pow / (num_input_channels * kFftSize); + } + + private: + std::unique_ptr transform_; + float window_[kFftSize]; + float low_pow_; + float high_pow_; +}; + +float EncodedPowerRatio(AudioEncoder* encoder, + AudioDecoder* decoder, + test::AudioLoop* audio_loop) { + // Encode and decode. + uint32_t rtp_timestamp = 0u; + constexpr size_t kBufferSize = 500; + rtc::Buffer encoded(kBufferSize); + std::vector decoded(kOutputBlockSizeSamples); + std::vector decoded_float(kOutputBlockSizeSamples); + AudioDecoder::SpeechType speech_type = AudioDecoder::kSpeech; + PowerRatioEstimator power_ratio_estimator; + for (size_t i = 0; i < 1000; ++i) { + encoded.Clear(); + AudioEncoder::EncodedInfo encoder_info = + encoder->Encode(rtp_timestamp, audio_loop->GetNextBlock(), &encoded); + rtp_timestamp += kInputBlockSizeSamples; + if (encoded.size() > 0) { + int decoder_info = decoder->Decode( + encoded.data(), encoded.size(), kSampleRateHz, + decoded.size() * sizeof(decoded[0]), decoded.data(), &speech_type); + if (decoder_info > 0) { + S16ToFloat(decoded.data(), decoded.size(), decoded_float.data()); + power_ratio_estimator.ProcessBlock(decoded_float.data()); + } + } + } + return power_ratio_estimator.PowerRatio(); +} + +} // namespace + +// TODO(ivoc): Remove this test, WebRTC-AdjustOpusBandwidth is obsolete. +TEST(BandwidthAdaptationTest, BandwidthAdaptationTest) { + test::ScopedFieldTrials override_field_trials( + "WebRTC-AdjustOpusBandwidth/Enabled/"); + + constexpr float kMaxNarrowbandRatio = 0.0035f; + constexpr float kMinWidebandRatio = 0.01f; + + // Create encoder. + AudioEncoderOpusConfig enc_config; + enc_config.bitrate_bps = absl::optional(7999); + enc_config.num_channels = kNumChannels; + constexpr int payload_type = 17; + auto encoder = AudioEncoderOpus::MakeAudioEncoder(enc_config, payload_type); + + // Create decoder. + AudioDecoderOpus::Config dec_config; + dec_config.num_channels = kNumChannels; + auto decoder = AudioDecoderOpus::MakeAudioDecoder(dec_config); + + // Open speech file. + const std::string kInputFileName = + webrtc::test::ResourcePath("audio_coding/speech_mono_32_48kHz", "pcm"); + test::AudioLoop audio_loop; + EXPECT_EQ(kSampleRateHz, encoder->SampleRateHz()); + ASSERT_TRUE(audio_loop.Init(kInputFileName, kMaxLoopLengthSamples, + kInputBlockSizeSamples)); + + EXPECT_LT(EncodedPowerRatio(encoder.get(), decoder.get(), &audio_loop), + kMaxNarrowbandRatio); + + encoder->OnReceivedTargetAudioBitrate(9000); + EXPECT_LT(EncodedPowerRatio(encoder.get(), decoder.get(), &audio_loop), + kMaxNarrowbandRatio); + + encoder->OnReceivedTargetAudioBitrate(9001); + EXPECT_GT(EncodedPowerRatio(encoder.get(), decoder.get(), &audio_loop), + kMinWidebandRatio); + + encoder->OnReceivedTargetAudioBitrate(8000); + EXPECT_GT(EncodedPowerRatio(encoder.get(), decoder.get(), &audio_loop), + kMinWidebandRatio); + + encoder->OnReceivedTargetAudioBitrate(12001); + EXPECT_GT(EncodedPowerRatio(encoder.get(), decoder.get(), &audio_loop), + kMinWidebandRatio); +} + +} // namespace webrtc diff --git a/third_party/libwebrtc/modules/audio_coding/codecs/opus/opus_complexity_unittest.cc b/third_party/libwebrtc/modules/audio_coding/codecs/opus/opus_complexity_unittest.cc new file mode 100644 index 0000000000..e8c131092c --- /dev/null +++ b/third_party/libwebrtc/modules/audio_coding/codecs/opus/opus_complexity_unittest.cc @@ -0,0 +1,105 @@ +/* + * Copyright (c) 2016 The WebRTC project authors. All Rights Reserved. + * + * Use of this source code is governed by a BSD-style license + * that can be found in the LICENSE file in the root of the source + * tree. An additional intellectual property rights grant can be found + * in the file PATENTS. All contributing project authors may + * be found in the AUTHORS file in the root of the source tree. + */ + +#include "api/audio_codecs/opus/audio_encoder_opus.h" +#include "api/test/metrics/global_metrics_logger_and_exporter.h" +#include "api/test/metrics/metric.h" +#include "modules/audio_coding/neteq/tools/audio_loop.h" +#include "rtc_base/time_utils.h" +#include "test/gtest.h" +#include "test/testsupport/file_utils.h" + +namespace webrtc { +namespace { + +using ::webrtc::test::GetGlobalMetricsLogger; +using ::webrtc::test::ImprovementDirection; +using ::webrtc::test::Unit; + +int64_t RunComplexityTest(const AudioEncoderOpusConfig& config) { + // Create encoder. + constexpr int payload_type = 17; + const auto encoder = AudioEncoderOpus::MakeAudioEncoder(config, payload_type); + // Open speech file. + const std::string kInputFileName = + webrtc::test::ResourcePath("audio_coding/speech_mono_32_48kHz", "pcm"); + test::AudioLoop audio_loop; + constexpr int kSampleRateHz = 48000; + EXPECT_EQ(kSampleRateHz, encoder->SampleRateHz()); + constexpr size_t kMaxLoopLengthSamples = + kSampleRateHz * 10; // 10 second loop. + constexpr size_t kInputBlockSizeSamples = + 10 * kSampleRateHz / 1000; // 60 ms. + EXPECT_TRUE(audio_loop.Init(kInputFileName, kMaxLoopLengthSamples, + kInputBlockSizeSamples)); + // Encode. + const int64_t start_time_ms = rtc::TimeMillis(); + AudioEncoder::EncodedInfo info; + rtc::Buffer encoded(500); + uint32_t rtp_timestamp = 0u; + for (size_t i = 0; i < 10000; ++i) { + encoded.Clear(); + info = encoder->Encode(rtp_timestamp, audio_loop.GetNextBlock(), &encoded); + rtp_timestamp += kInputBlockSizeSamples; + } + return rtc::TimeMillis() - start_time_ms; +} + +// This test encodes an audio file using Opus twice with different bitrates +// (~11 kbps and 15.5 kbps). The runtime for each is measured, and the ratio +// between the two is calculated and tracked. This test explicitly sets the +// low_rate_complexity to 9. When running on desktop platforms, this is the same +// as the regular complexity, and the expectation is that the resulting ratio +// should be less than 100% (since the encoder runs faster at lower bitrates, +// given a fixed complexity setting). On the other hand, when running on +// mobiles, the regular complexity is 5, and we expect the resulting ratio to +// be higher, since we have explicitly asked for a higher complexity setting at +// the lower rate. +TEST(AudioEncoderOpusComplexityAdaptationTest, Adaptation_On) { + // Create config. + AudioEncoderOpusConfig config; + // The limit -- including the hysteresis window -- at which the complexity + // shuold be increased. + config.bitrate_bps = 11000 - 1; + config.low_rate_complexity = 9; + int64_t runtime_10999bps = RunComplexityTest(config); + + config.bitrate_bps = 15500; + int64_t runtime_15500bps = RunComplexityTest(config); + + GetGlobalMetricsLogger()->LogSingleValueMetric( + "opus_encoding_complexity_ratio", "adaptation_on", + 100.0 * runtime_10999bps / runtime_15500bps, Unit::kPercent, + ImprovementDirection::kNeitherIsBetter); +} + +// This test is identical to the one above, but without the complexity +// adaptation enabled (neither on desktop, nor on mobile). The expectation is +// that the resulting ratio is less than 100% at all times. +TEST(AudioEncoderOpusComplexityAdaptationTest, Adaptation_Off) { + // Create config. + AudioEncoderOpusConfig config; + // The limit -- including the hysteresis window -- at which the complexity + // shuold be increased (but not in this test since complexity adaptation is + // disabled). + config.bitrate_bps = 11000 - 1; + int64_t runtime_10999bps = RunComplexityTest(config); + + config.bitrate_bps = 15500; + int64_t runtime_15500bps = RunComplexityTest(config); + + GetGlobalMetricsLogger()->LogSingleValueMetric( + "opus_encoding_complexity_ratio", "adaptation_off", + 100.0 * runtime_10999bps / runtime_15500bps, Unit::kPercent, + ImprovementDirection::kNeitherIsBetter); +} + +} // namespace +} // namespace webrtc diff --git a/third_party/libwebrtc/modules/audio_coding/codecs/opus/opus_fec_test.cc b/third_party/libwebrtc/modules/audio_coding/codecs/opus/opus_fec_test.cc new file mode 100644 index 0000000000..815f26e31c --- /dev/null +++ b/third_party/libwebrtc/modules/audio_coding/codecs/opus/opus_fec_test.cc @@ -0,0 +1,248 @@ +/* + * Copyright (c) 2014 The WebRTC project authors. All Rights Reserved. + * + * Use of this source code is governed by a BSD-style license + * that can be found in the LICENSE file in the root of the source + * tree. An additional intellectual property rights grant can be found + * in the file PATENTS. All contributing project authors may + * be found in the AUTHORS file in the root of the source tree. + */ + +#include + +#include "modules/audio_coding/codecs/opus/opus_interface.h" +#include "test/gtest.h" +#include "test/testsupport/file_utils.h" + +using std::get; +using std::string; +using std::tuple; +using ::testing::TestWithParam; + +namespace webrtc { + +// Define coding parameter as . +typedef tuple coding_param; +typedef struct mode mode; + +struct mode { + bool fec; + uint8_t target_packet_loss_rate; +}; + +const int kOpusBlockDurationMs = 20; +const int kOpusSamplingKhz = 48; + +class OpusFecTest : public TestWithParam { + protected: + OpusFecTest(); + + void SetUp() override; + void TearDown() override; + + virtual void EncodeABlock(); + + virtual void DecodeABlock(bool lost_previous, bool lost_current); + + int block_duration_ms_; + int sampling_khz_; + size_t block_length_sample_; + + size_t channels_; + int bit_rate_; + + size_t data_pointer_; + size_t loop_length_samples_; + size_t max_bytes_; + size_t encoded_bytes_; + + WebRtcOpusEncInst* opus_encoder_; + WebRtcOpusDecInst* opus_decoder_; + + string in_filename_; + + std::unique_ptr in_data_; + std::unique_ptr out_data_; + std::unique_ptr bit_stream_; +}; + +void OpusFecTest::SetUp() { + channels_ = get<0>(GetParam()); + bit_rate_ = get<1>(GetParam()); + printf("Coding %zu channel signal at %d bps.\n", channels_, bit_rate_); + + in_filename_ = test::ResourcePath(get<2>(GetParam()), get<3>(GetParam())); + + FILE* fp = fopen(in_filename_.c_str(), "rb"); + ASSERT_FALSE(fp == NULL); + + // Obtain file size. + fseek(fp, 0, SEEK_END); + loop_length_samples_ = ftell(fp) / sizeof(int16_t); + rewind(fp); + + // Allocate memory to contain the whole file. + in_data_.reset( + new int16_t[loop_length_samples_ + block_length_sample_ * channels_]); + + // Copy the file into the buffer. + ASSERT_EQ(fread(&in_data_[0], sizeof(int16_t), loop_length_samples_, fp), + loop_length_samples_); + fclose(fp); + + // The audio will be used in a looped manner. To ease the acquisition of an + // audio frame that crosses the end of the excerpt, we add an extra block + // length of samples to the end of the array, starting over again from the + // beginning of the array. Audio frames cross the end of the excerpt always + // appear as a continuum of memory. + memcpy(&in_data_[loop_length_samples_], &in_data_[0], + block_length_sample_ * channels_ * sizeof(int16_t)); + + // Maximum number of bytes in output bitstream. + max_bytes_ = block_length_sample_ * channels_ * sizeof(int16_t); + + out_data_.reset(new int16_t[2 * block_length_sample_ * channels_]); + bit_stream_.reset(new uint8_t[max_bytes_]); + + // If channels_ == 1, use Opus VOIP mode, otherwise, audio mode. + int app = channels_ == 1 ? 0 : 1; + + // Create encoder memory. + EXPECT_EQ(0, WebRtcOpus_EncoderCreate(&opus_encoder_, channels_, app, 48000)); + EXPECT_EQ(0, WebRtcOpus_DecoderCreate(&opus_decoder_, channels_, 48000)); + // Set bitrate. + EXPECT_EQ(0, WebRtcOpus_SetBitRate(opus_encoder_, bit_rate_)); +} + +void OpusFecTest::TearDown() { + // Free memory. + EXPECT_EQ(0, WebRtcOpus_EncoderFree(opus_encoder_)); + EXPECT_EQ(0, WebRtcOpus_DecoderFree(opus_decoder_)); +} + +OpusFecTest::OpusFecTest() + : block_duration_ms_(kOpusBlockDurationMs), + sampling_khz_(kOpusSamplingKhz), + block_length_sample_( + static_cast(block_duration_ms_ * sampling_khz_)), + data_pointer_(0), + max_bytes_(0), + encoded_bytes_(0), + opus_encoder_(NULL), + opus_decoder_(NULL) {} + +void OpusFecTest::EncodeABlock() { + int value = + WebRtcOpus_Encode(opus_encoder_, &in_data_[data_pointer_], + block_length_sample_, max_bytes_, &bit_stream_[0]); + EXPECT_GT(value, 0); + + encoded_bytes_ = static_cast(value); +} + +void OpusFecTest::DecodeABlock(bool lost_previous, bool lost_current) { + int16_t audio_type; + int value_1 = 0, value_2 = 0; + + if (lost_previous) { + // Decode previous frame. + if (!lost_current && + WebRtcOpus_PacketHasFec(&bit_stream_[0], encoded_bytes_) == 1) { + value_1 = + WebRtcOpus_DecodeFec(opus_decoder_, &bit_stream_[0], encoded_bytes_, + &out_data_[0], &audio_type); + } else { + // Call decoder PLC. + while (value_1 < static_cast(block_length_sample_)) { + int ret = WebRtcOpus_Decode(opus_decoder_, NULL, 0, &out_data_[value_1], + &audio_type); + EXPECT_EQ(ret, sampling_khz_ * 10); // Should return 10 ms of samples. + value_1 += ret; + } + } + EXPECT_EQ(static_cast(block_length_sample_), value_1); + } + + if (!lost_current) { + // Decode current frame. + value_2 = WebRtcOpus_Decode(opus_decoder_, &bit_stream_[0], encoded_bytes_, + &out_data_[value_1 * channels_], &audio_type); + EXPECT_EQ(static_cast(block_length_sample_), value_2); + } +} + +TEST_P(OpusFecTest, RandomPacketLossTest) { + const int kDurationMs = 200000; + int time_now_ms, fec_frames; + int actual_packet_loss_rate; + bool lost_current, lost_previous; + mode mode_set[3] = {{true, 0}, {false, 0}, {true, 50}}; + + lost_current = false; + for (int i = 0; i < 3; i++) { + if (mode_set[i].fec) { + EXPECT_EQ(0, WebRtcOpus_EnableFec(opus_encoder_)); + EXPECT_EQ(0, WebRtcOpus_SetPacketLossRate( + opus_encoder_, mode_set[i].target_packet_loss_rate)); + printf("FEC is ON, target at packet loss rate %d percent.\n", + mode_set[i].target_packet_loss_rate); + } else { + EXPECT_EQ(0, WebRtcOpus_DisableFec(opus_encoder_)); + printf("FEC is OFF.\n"); + } + // In this test, we let the target packet loss rate match the actual rate. + actual_packet_loss_rate = mode_set[i].target_packet_loss_rate; + // Run every mode a certain time. + time_now_ms = 0; + fec_frames = 0; + while (time_now_ms < kDurationMs) { + // Encode & decode. + EncodeABlock(); + + // Check if payload has FEC. + int fec = WebRtcOpus_PacketHasFec(&bit_stream_[0], encoded_bytes_); + + // If FEC is disabled or the target packet loss rate is set to 0, there + // should be no FEC in the bit stream. + if (!mode_set[i].fec || mode_set[i].target_packet_loss_rate == 0) { + EXPECT_EQ(fec, 0); + } else if (fec == 1) { + fec_frames++; + } + + lost_previous = lost_current; + lost_current = rand() < actual_packet_loss_rate * (RAND_MAX / 100); + DecodeABlock(lost_previous, lost_current); + + time_now_ms += block_duration_ms_; + + // `data_pointer_` is incremented and wrapped across + // `loop_length_samples_`. + data_pointer_ = (data_pointer_ + block_length_sample_ * channels_) % + loop_length_samples_; + } + if (mode_set[i].fec) { + printf("%.2f percent frames has FEC.\n", + static_cast(fec_frames) * block_duration_ms_ / 2000); + } + } +} + +const coding_param param_set[] = { + std::make_tuple(1, + 64000, + string("audio_coding/testfile32kHz"), + string("pcm")), + std::make_tuple(1, + 32000, + string("audio_coding/testfile32kHz"), + string("pcm")), + std::make_tuple(2, + 64000, + string("audio_coding/teststereo32kHz"), + string("pcm"))}; + +// 64 kbps, stereo +INSTANTIATE_TEST_SUITE_P(AllTest, OpusFecTest, ::testing::ValuesIn(param_set)); + +} // namespace webrtc diff --git a/third_party/libwebrtc/modules/audio_coding/codecs/opus/opus_inst.h b/third_party/libwebrtc/modules/audio_coding/codecs/opus/opus_inst.h new file mode 100644 index 0000000000..92c5c354a7 --- /dev/null +++ b/third_party/libwebrtc/modules/audio_coding/codecs/opus/opus_inst.h @@ -0,0 +1,43 @@ +/* + * Copyright (c) 2014 The WebRTC project authors. All Rights Reserved. + * + * Use of this source code is governed by a BSD-style license + * that can be found in the LICENSE file in the root of the source + * tree. An additional intellectual property rights grant can be found + * in the file PATENTS. All contributing project authors may + * be found in the AUTHORS file in the root of the source tree. + */ + +#ifndef MODULES_AUDIO_CODING_CODECS_OPUS_OPUS_INST_H_ +#define MODULES_AUDIO_CODING_CODECS_OPUS_OPUS_INST_H_ + +#include + +#include "rtc_base/ignore_wundef.h" + +RTC_PUSH_IGNORING_WUNDEF() +#include "third_party/opus/src/include/opus.h" +#include "third_party/opus/src/include/opus_multistream.h" +RTC_POP_IGNORING_WUNDEF() + +struct WebRtcOpusEncInst { + OpusEncoder* encoder; + OpusMSEncoder* multistream_encoder; + size_t channels; + int in_dtx_mode; + bool avoid_noise_pumping_during_dtx; + int sample_rate_hz; + float smooth_energy_non_active_frames; +}; + +struct WebRtcOpusDecInst { + OpusDecoder* decoder; + OpusMSDecoder* multistream_decoder; + int prev_decoded_samples; + bool plc_use_prev_decoded_samples; + size_t channels; + int in_dtx_mode; + int sample_rate_hz; +}; + +#endif // MODULES_AUDIO_CODING_CODECS_OPUS_OPUS_INST_H_ diff --git a/third_party/libwebrtc/modules/audio_coding/codecs/opus/opus_interface.cc b/third_party/libwebrtc/modules/audio_coding/codecs/opus/opus_interface.cc new file mode 100644 index 0000000000..67d8619b34 --- /dev/null +++ b/third_party/libwebrtc/modules/audio_coding/codecs/opus/opus_interface.cc @@ -0,0 +1,881 @@ +/* + * Copyright (c) 2012 The WebRTC project authors. All Rights Reserved. + * + * Use of this source code is governed by a BSD-style license + * that can be found in the LICENSE file in the root of the source + * tree. An additional intellectual property rights grant can be found + * in the file PATENTS. All contributing project authors may + * be found in the AUTHORS file in the root of the source tree. + */ + +#include "modules/audio_coding/codecs/opus/opus_interface.h" + +#include + +#include + +#include "api/array_view.h" +#include "rtc_base/checks.h" +#include "system_wrappers/include/field_trial.h" + +enum { +#if WEBRTC_OPUS_SUPPORT_120MS_PTIME + /* Maximum supported frame size in WebRTC is 120 ms. */ + kWebRtcOpusMaxEncodeFrameSizeMs = 120, +#else + /* Maximum supported frame size in WebRTC is 60 ms. */ + kWebRtcOpusMaxEncodeFrameSizeMs = 60, +#endif + + /* The format allows up to 120 ms frames. Since we don't control the other + * side, we must allow for packets of that size. NetEq is currently limited + * to 60 ms on the receive side. */ + kWebRtcOpusMaxDecodeFrameSizeMs = 120, + + // Duration of audio that each call to packet loss concealment covers. + kWebRtcOpusPlcFrameSizeMs = 10, +}; + +constexpr char kPlcUsePrevDecodedSamplesFieldTrial[] = + "WebRTC-Audio-OpusPlcUsePrevDecodedSamples"; + +constexpr char kAvoidNoisePumpingDuringDtxFieldTrial[] = + "WebRTC-Audio-OpusAvoidNoisePumpingDuringDtx"; + +constexpr char kSetSignalVoiceWithDtxFieldTrial[] = + "WebRTC-Audio-OpusSetSignalVoiceWithDtx"; + +static int FrameSizePerChannel(int frame_size_ms, int sample_rate_hz) { + RTC_DCHECK_GT(frame_size_ms, 0); + RTC_DCHECK_EQ(frame_size_ms % 10, 0); + RTC_DCHECK_GT(sample_rate_hz, 0); + RTC_DCHECK_EQ(sample_rate_hz % 1000, 0); + return frame_size_ms * (sample_rate_hz / 1000); +} + +// Maximum sample count per channel. +static int MaxFrameSizePerChannel(int sample_rate_hz) { + return FrameSizePerChannel(kWebRtcOpusMaxDecodeFrameSizeMs, sample_rate_hz); +} + +// Default sample count per channel. +static int DefaultFrameSizePerChannel(int sample_rate_hz) { + return FrameSizePerChannel(20, sample_rate_hz); +} + +// Returns true if the `encoded` payload corresponds to a refresh DTX packet +// whose energy is larger than the expected for non activity packets. +static bool WebRtcOpus_IsHighEnergyRefreshDtxPacket( + OpusEncInst* inst, + rtc::ArrayView frame, + rtc::ArrayView encoded) { + if (encoded.size() <= 2) { + return false; + } + int number_frames = + frame.size() / DefaultFrameSizePerChannel(inst->sample_rate_hz); + if (number_frames > 0 && + WebRtcOpus_PacketHasVoiceActivity(encoded.data(), encoded.size()) == 0) { + const float average_frame_energy = + std::accumulate(frame.begin(), frame.end(), 0.0f, + [](float a, int32_t b) { return a + b * b; }) / + number_frames; + if (WebRtcOpus_GetInDtx(inst) == 1 && + average_frame_energy >= inst->smooth_energy_non_active_frames * 0.5f) { + // This is a refresh DTX packet as the encoder is in DTX and has + // produced a payload > 2 bytes. This refresh packet has a higher energy + // than the smooth energy of non activity frames (with a 3 dB negative + // margin) and, therefore, it is flagged as a high energy refresh DTX + // packet. + return true; + } + // The average energy is tracked in a similar way as the modeling of the + // comfort noise in the Silk decoder in Opus + // (third_party/opus/src/silk/CNG.c). + if (average_frame_energy < inst->smooth_energy_non_active_frames * 0.5f) { + inst->smooth_energy_non_active_frames = average_frame_energy; + } else { + inst->smooth_energy_non_active_frames += + (average_frame_energy - inst->smooth_energy_non_active_frames) * + 0.25f; + } + } + return false; +} + +int16_t WebRtcOpus_EncoderCreate(OpusEncInst** inst, + size_t channels, + int32_t application, + int sample_rate_hz) { + int opus_app; + if (!inst) + return -1; + + switch (application) { + case 0: + opus_app = OPUS_APPLICATION_VOIP; + break; + case 1: + opus_app = OPUS_APPLICATION_AUDIO; + break; + default: + return -1; + } + + OpusEncInst* state = + reinterpret_cast(calloc(1, sizeof(OpusEncInst))); + RTC_DCHECK(state); + + int error; + state->encoder = opus_encoder_create( + sample_rate_hz, static_cast(channels), opus_app, &error); + + if (error != OPUS_OK || (!state->encoder && !state->multistream_encoder)) { + WebRtcOpus_EncoderFree(state); + return -1; + } + + state->in_dtx_mode = 0; + state->channels = channels; + state->sample_rate_hz = sample_rate_hz; + state->smooth_energy_non_active_frames = 0.0f; + state->avoid_noise_pumping_during_dtx = + webrtc::field_trial::IsEnabled(kAvoidNoisePumpingDuringDtxFieldTrial); + + *inst = state; + return 0; +} + +int16_t WebRtcOpus_MultistreamEncoderCreate( + OpusEncInst** inst, + size_t channels, + int32_t application, + size_t streams, + size_t coupled_streams, + const unsigned char* channel_mapping) { + int opus_app; + if (!inst) + return -1; + + switch (application) { + case 0: + opus_app = OPUS_APPLICATION_VOIP; + break; + case 1: + opus_app = OPUS_APPLICATION_AUDIO; + break; + default: + return -1; + } + + OpusEncInst* state = + reinterpret_cast(calloc(1, sizeof(OpusEncInst))); + RTC_DCHECK(state); + + int error; + const int sample_rate_hz = 48000; + state->multistream_encoder = opus_multistream_encoder_create( + sample_rate_hz, channels, streams, coupled_streams, channel_mapping, + opus_app, &error); + + if (error != OPUS_OK || (!state->encoder && !state->multistream_encoder)) { + WebRtcOpus_EncoderFree(state); + return -1; + } + + state->in_dtx_mode = 0; + state->channels = channels; + state->sample_rate_hz = sample_rate_hz; + state->smooth_energy_non_active_frames = 0.0f; + state->avoid_noise_pumping_during_dtx = false; + + *inst = state; + return 0; +} + +int16_t WebRtcOpus_EncoderFree(OpusEncInst* inst) { + if (inst) { + if (inst->encoder) { + opus_encoder_destroy(inst->encoder); + } else { + opus_multistream_encoder_destroy(inst->multistream_encoder); + } + free(inst); + return 0; + } else { + return -1; + } +} + +int WebRtcOpus_Encode(OpusEncInst* inst, + const int16_t* audio_in, + size_t samples, + size_t length_encoded_buffer, + uint8_t* encoded) { + int res; + + if (samples > 48 * kWebRtcOpusMaxEncodeFrameSizeMs) { + return -1; + } + + if (inst->encoder) { + res = opus_encode(inst->encoder, (const opus_int16*)audio_in, + static_cast(samples), encoded, + static_cast(length_encoded_buffer)); + } else { + res = opus_multistream_encode( + inst->multistream_encoder, (const opus_int16*)audio_in, + static_cast(samples), encoded, + static_cast(length_encoded_buffer)); + } + + if (res <= 0) { + return -1; + } + + if (res <= 2) { + // Indicates DTX since the packet has nothing but a header. In principle, + // there is no need to send this packet. However, we do transmit the first + // occurrence to let the decoder know that the encoder enters DTX mode. + if (inst->in_dtx_mode) { + return 0; + } else { + inst->in_dtx_mode = 1; + return res; + } + } + + if (inst->avoid_noise_pumping_during_dtx && WebRtcOpus_GetUseDtx(inst) == 1 && + WebRtcOpus_IsHighEnergyRefreshDtxPacket( + inst, rtc::MakeArrayView(audio_in, samples), + rtc::MakeArrayView(encoded, res))) { + // This packet is a high energy refresh DTX packet. For avoiding an increase + // of the energy in the DTX region at the decoder, this packet is + // substituted by a TOC byte with one empty frame. + // The number of frames described in the TOC byte + // (https://tools.ietf.org/html/rfc6716#section-3.1) are overwritten to + // always indicate one frame (last two bits equal to 0). + encoded[0] = encoded[0] & 0b11111100; + inst->in_dtx_mode = 1; + // The payload is just the TOC byte and has 1 byte as length. + return 1; + } + inst->in_dtx_mode = 0; + return res; +} + +#define ENCODER_CTL(inst, vargs) \ + (inst->encoder \ + ? opus_encoder_ctl(inst->encoder, vargs) \ + : opus_multistream_encoder_ctl(inst->multistream_encoder, vargs)) + +int16_t WebRtcOpus_SetBitRate(OpusEncInst* inst, int32_t rate) { + if (inst) { + return ENCODER_CTL(inst, OPUS_SET_BITRATE(rate)); + } else { + return -1; + } +} + +int16_t WebRtcOpus_SetPacketLossRate(OpusEncInst* inst, int32_t loss_rate) { + if (inst) { + return ENCODER_CTL(inst, OPUS_SET_PACKET_LOSS_PERC(loss_rate)); + } else { + return -1; + } +} + +int16_t WebRtcOpus_SetMaxPlaybackRate(OpusEncInst* inst, int32_t frequency_hz) { + opus_int32 set_bandwidth; + + if (!inst) + return -1; + + if (frequency_hz <= 8000) { + set_bandwidth = OPUS_BANDWIDTH_NARROWBAND; + } else if (frequency_hz <= 12000) { + set_bandwidth = OPUS_BANDWIDTH_MEDIUMBAND; + } else if (frequency_hz <= 16000) { + set_bandwidth = OPUS_BANDWIDTH_WIDEBAND; + } else if (frequency_hz <= 24000) { + set_bandwidth = OPUS_BANDWIDTH_SUPERWIDEBAND; + } else { + set_bandwidth = OPUS_BANDWIDTH_FULLBAND; + } + return ENCODER_CTL(inst, OPUS_SET_MAX_BANDWIDTH(set_bandwidth)); +} + +int16_t WebRtcOpus_GetMaxPlaybackRate(OpusEncInst* const inst, + int32_t* result_hz) { + if (inst->encoder) { + if (opus_encoder_ctl(inst->encoder, OPUS_GET_MAX_BANDWIDTH(result_hz)) == + OPUS_OK) { + return 0; + } + return -1; + } + + opus_int32 max_bandwidth; + int s; + int ret; + + max_bandwidth = 0; + ret = OPUS_OK; + s = 0; + while (ret == OPUS_OK) { + OpusEncoder* enc; + opus_int32 bandwidth; + + ret = ENCODER_CTL(inst, OPUS_MULTISTREAM_GET_ENCODER_STATE(s, &enc)); + if (ret == OPUS_BAD_ARG) + break; + if (ret != OPUS_OK) + return -1; + if (opus_encoder_ctl(enc, OPUS_GET_MAX_BANDWIDTH(&bandwidth)) != OPUS_OK) + return -1; + + if (max_bandwidth != 0 && max_bandwidth != bandwidth) + return -1; + + max_bandwidth = bandwidth; + s++; + } + *result_hz = max_bandwidth; + return 0; +} + +int16_t WebRtcOpus_EnableFec(OpusEncInst* inst) { + if (inst) { + return ENCODER_CTL(inst, OPUS_SET_INBAND_FEC(1)); + } else { + return -1; + } +} + +int16_t WebRtcOpus_DisableFec(OpusEncInst* inst) { + if (inst) { + return ENCODER_CTL(inst, OPUS_SET_INBAND_FEC(0)); + } else { + return -1; + } +} + +int16_t WebRtcOpus_EnableDtx(OpusEncInst* inst) { + if (inst) { + if (webrtc::field_trial::IsEnabled(kSetSignalVoiceWithDtxFieldTrial)) { + int ret = ENCODER_CTL(inst, OPUS_SET_SIGNAL(OPUS_SIGNAL_VOICE)); + if (ret != OPUS_OK) { + return ret; + } + } + return ENCODER_CTL(inst, OPUS_SET_DTX(1)); + } else { + return -1; + } +} + +int16_t WebRtcOpus_DisableDtx(OpusEncInst* inst) { + if (inst) { + if (webrtc::field_trial::IsEnabled(kSetSignalVoiceWithDtxFieldTrial)) { + int ret = ENCODER_CTL(inst, OPUS_SET_SIGNAL(OPUS_AUTO)); + if (ret != OPUS_OK) { + return ret; + } + } + return ENCODER_CTL(inst, OPUS_SET_DTX(0)); + } else { + return -1; + } +} + +int16_t WebRtcOpus_GetUseDtx(OpusEncInst* inst) { + if (inst) { + opus_int32 use_dtx; + if (ENCODER_CTL(inst, OPUS_GET_DTX(&use_dtx)) == 0) { + return use_dtx; + } + } + return -1; +} + +int16_t WebRtcOpus_EnableCbr(OpusEncInst* inst) { + if (inst) { + return ENCODER_CTL(inst, OPUS_SET_VBR(0)); + } else { + return -1; + } +} + +int16_t WebRtcOpus_DisableCbr(OpusEncInst* inst) { + if (inst) { + return ENCODER_CTL(inst, OPUS_SET_VBR(1)); + } else { + return -1; + } +} + +int16_t WebRtcOpus_SetComplexity(OpusEncInst* inst, int32_t complexity) { + if (inst) { + return ENCODER_CTL(inst, OPUS_SET_COMPLEXITY(complexity)); + } else { + return -1; + } +} + +int32_t WebRtcOpus_GetBandwidth(OpusEncInst* inst) { + if (!inst) { + return -1; + } + int32_t bandwidth; + if (ENCODER_CTL(inst, OPUS_GET_BANDWIDTH(&bandwidth)) == 0) { + return bandwidth; + } else { + return -1; + } +} + +int16_t WebRtcOpus_SetBandwidth(OpusEncInst* inst, int32_t bandwidth) { + if (inst) { + return ENCODER_CTL(inst, OPUS_SET_BANDWIDTH(bandwidth)); + } else { + return -1; + } +} + +int16_t WebRtcOpus_SetForceChannels(OpusEncInst* inst, size_t num_channels) { + if (!inst) + return -1; + if (num_channels == 0) { + return ENCODER_CTL(inst, OPUS_SET_FORCE_CHANNELS(OPUS_AUTO)); + } else if (num_channels == 1 || num_channels == 2) { + return ENCODER_CTL(inst, OPUS_SET_FORCE_CHANNELS(num_channels)); + } else { + return -1; + } +} + +int32_t WebRtcOpus_GetInDtx(OpusEncInst* inst) { + if (!inst) { + return -1; + } +#ifdef OPUS_GET_IN_DTX + int32_t in_dtx; + if (ENCODER_CTL(inst, OPUS_GET_IN_DTX(&in_dtx)) == 0) { + return in_dtx; + } +#endif + return -1; +} + +int16_t WebRtcOpus_DecoderCreate(OpusDecInst** inst, + size_t channels, + int sample_rate_hz) { + int error; + OpusDecInst* state; + + if (inst != NULL) { + // Create Opus decoder state. + state = reinterpret_cast(calloc(1, sizeof(OpusDecInst))); + if (state == NULL) { + return -1; + } + + state->decoder = + opus_decoder_create(sample_rate_hz, static_cast(channels), &error); + if (error == OPUS_OK && state->decoder) { + // Creation of memory all ok. + state->channels = channels; + state->sample_rate_hz = sample_rate_hz; + state->plc_use_prev_decoded_samples = + webrtc::field_trial::IsEnabled(kPlcUsePrevDecodedSamplesFieldTrial); + if (state->plc_use_prev_decoded_samples) { + state->prev_decoded_samples = + DefaultFrameSizePerChannel(state->sample_rate_hz); + } + state->in_dtx_mode = 0; + *inst = state; + return 0; + } + + // If memory allocation was unsuccessful, free the entire state. + if (state->decoder) { + opus_decoder_destroy(state->decoder); + } + free(state); + } + return -1; +} + +int16_t WebRtcOpus_MultistreamDecoderCreate( + OpusDecInst** inst, + size_t channels, + size_t streams, + size_t coupled_streams, + const unsigned char* channel_mapping) { + int error; + OpusDecInst* state; + + if (inst != NULL) { + // Create Opus decoder state. + state = reinterpret_cast(calloc(1, sizeof(OpusDecInst))); + if (state == NULL) { + return -1; + } + + // Create new memory, always at 48000 Hz. + state->multistream_decoder = opus_multistream_decoder_create( + 48000, channels, streams, coupled_streams, channel_mapping, &error); + + if (error == OPUS_OK && state->multistream_decoder) { + // Creation of memory all ok. + state->channels = channels; + state->sample_rate_hz = 48000; + state->plc_use_prev_decoded_samples = + webrtc::field_trial::IsEnabled(kPlcUsePrevDecodedSamplesFieldTrial); + if (state->plc_use_prev_decoded_samples) { + state->prev_decoded_samples = + DefaultFrameSizePerChannel(state->sample_rate_hz); + } + state->in_dtx_mode = 0; + *inst = state; + return 0; + } + + // If memory allocation was unsuccessful, free the entire state. + opus_multistream_decoder_destroy(state->multistream_decoder); + free(state); + } + return -1; +} + +int16_t WebRtcOpus_DecoderFree(OpusDecInst* inst) { + if (inst) { + if (inst->decoder) { + opus_decoder_destroy(inst->decoder); + } else if (inst->multistream_decoder) { + opus_multistream_decoder_destroy(inst->multistream_decoder); + } + free(inst); + return 0; + } else { + return -1; + } +} + +size_t WebRtcOpus_DecoderChannels(OpusDecInst* inst) { + return inst->channels; +} + +void WebRtcOpus_DecoderInit(OpusDecInst* inst) { + if (inst->decoder) { + opus_decoder_ctl(inst->decoder, OPUS_RESET_STATE); + } else { + opus_multistream_decoder_ctl(inst->multistream_decoder, OPUS_RESET_STATE); + } + inst->in_dtx_mode = 0; +} + +/* For decoder to determine if it is to output speech or comfort noise. */ +static int16_t DetermineAudioType(OpusDecInst* inst, size_t encoded_bytes) { + // Audio type becomes comfort noise if `encoded_byte` is 1 and keeps + // to be so if the following `encoded_byte` are 0 or 1. + if (encoded_bytes == 0 && inst->in_dtx_mode) { + return 2; // Comfort noise. + } else if (encoded_bytes == 1 || encoded_bytes == 2) { + // TODO(henrik.lundin): There is a slight risk that a 2-byte payload is in + // fact a 1-byte TOC with a 1-byte payload. That will be erroneously + // interpreted as comfort noise output, but such a payload is probably + // faulty anyway. + + // TODO(webrtc:10218): This is wrong for multistream opus. Then are several + // single-stream packets glued together with some packet size bytes in + // between. See https://tools.ietf.org/html/rfc6716#appendix-B + inst->in_dtx_mode = 1; + return 2; // Comfort noise. + } else { + inst->in_dtx_mode = 0; + return 0; // Speech. + } +} + +/* `frame_size` is set to maximum Opus frame size in the normal case, and + * is set to the number of samples needed for PLC in case of losses. + * It is up to the caller to make sure the value is correct. */ +static int DecodeNative(OpusDecInst* inst, + const uint8_t* encoded, + size_t encoded_bytes, + int frame_size, + int16_t* decoded, + int16_t* audio_type, + int decode_fec) { + int res = -1; + if (inst->decoder) { + res = opus_decode( + inst->decoder, encoded, static_cast(encoded_bytes), + reinterpret_cast(decoded), frame_size, decode_fec); + } else { + res = opus_multistream_decode(inst->multistream_decoder, encoded, + static_cast(encoded_bytes), + reinterpret_cast(decoded), + frame_size, decode_fec); + } + + if (res <= 0) + return -1; + + *audio_type = DetermineAudioType(inst, encoded_bytes); + + return res; +} + +static int DecodePlc(OpusDecInst* inst, int16_t* decoded) { + int16_t audio_type = 0; + int decoded_samples; + int plc_samples = + FrameSizePerChannel(kWebRtcOpusPlcFrameSizeMs, inst->sample_rate_hz); + + if (inst->plc_use_prev_decoded_samples) { + /* The number of samples we ask for is `number_of_lost_frames` times + * `prev_decoded_samples_`. Limit the number of samples to maximum + * `MaxFrameSizePerChannel()`. */ + plc_samples = inst->prev_decoded_samples; + const int max_samples_per_channel = + MaxFrameSizePerChannel(inst->sample_rate_hz); + plc_samples = plc_samples <= max_samples_per_channel + ? plc_samples + : max_samples_per_channel; + } + decoded_samples = + DecodeNative(inst, NULL, 0, plc_samples, decoded, &audio_type, 0); + if (decoded_samples < 0) { + return -1; + } + + return decoded_samples; +} + +int WebRtcOpus_Decode(OpusDecInst* inst, + const uint8_t* encoded, + size_t encoded_bytes, + int16_t* decoded, + int16_t* audio_type) { + int decoded_samples; + + if (encoded_bytes == 0) { + *audio_type = DetermineAudioType(inst, encoded_bytes); + decoded_samples = DecodePlc(inst, decoded); + } else { + decoded_samples = DecodeNative(inst, encoded, encoded_bytes, + MaxFrameSizePerChannel(inst->sample_rate_hz), + decoded, audio_type, 0); + } + if (decoded_samples < 0) { + return -1; + } + + if (inst->plc_use_prev_decoded_samples) { + /* Update decoded sample memory, to be used by the PLC in case of losses. */ + inst->prev_decoded_samples = decoded_samples; + } + + return decoded_samples; +} + +int WebRtcOpus_DecodeFec(OpusDecInst* inst, + const uint8_t* encoded, + size_t encoded_bytes, + int16_t* decoded, + int16_t* audio_type) { + int decoded_samples; + int fec_samples; + + if (WebRtcOpus_PacketHasFec(encoded, encoded_bytes) != 1) { + return 0; + } + + fec_samples = + opus_packet_get_samples_per_frame(encoded, inst->sample_rate_hz); + + decoded_samples = DecodeNative(inst, encoded, encoded_bytes, fec_samples, + decoded, audio_type, 1); + if (decoded_samples < 0) { + return -1; + } + + return decoded_samples; +} + +int WebRtcOpus_DurationEst(OpusDecInst* inst, + const uint8_t* payload, + size_t payload_length_bytes) { + if (payload_length_bytes == 0) { + // WebRtcOpus_Decode calls PLC when payload length is zero. So we return + // PLC duration correspondingly. + return WebRtcOpus_PlcDuration(inst); + } + + int frames, samples; + frames = opus_packet_get_nb_frames( + payload, static_cast(payload_length_bytes)); + if (frames < 0) { + /* Invalid payload data. */ + return 0; + } + samples = + frames * opus_packet_get_samples_per_frame(payload, inst->sample_rate_hz); + if (samples > 120 * inst->sample_rate_hz / 1000) { + // More than 120 ms' worth of samples. + return 0; + } + return samples; +} + +int WebRtcOpus_PlcDuration(OpusDecInst* inst) { + if (inst->plc_use_prev_decoded_samples) { + /* The number of samples we ask for is `number_of_lost_frames` times + * `prev_decoded_samples_`. Limit the number of samples to maximum + * `MaxFrameSizePerChannel()`. */ + const int plc_samples = inst->prev_decoded_samples; + const int max_samples_per_channel = + MaxFrameSizePerChannel(inst->sample_rate_hz); + return plc_samples <= max_samples_per_channel ? plc_samples + : max_samples_per_channel; + } + return FrameSizePerChannel(kWebRtcOpusPlcFrameSizeMs, inst->sample_rate_hz); +} + +int WebRtcOpus_FecDurationEst(const uint8_t* payload, + size_t payload_length_bytes, + int sample_rate_hz) { + if (WebRtcOpus_PacketHasFec(payload, payload_length_bytes) != 1) { + return 0; + } + const int samples = + opus_packet_get_samples_per_frame(payload, sample_rate_hz); + const int samples_per_ms = sample_rate_hz / 1000; + if (samples < 10 * samples_per_ms || samples > 120 * samples_per_ms) { + /* Invalid payload duration. */ + return 0; + } + return samples; +} + +int WebRtcOpus_NumSilkFrames(const uint8_t* payload) { + // For computing the payload length in ms, the sample rate is not important + // since it cancels out. We use 48 kHz, but any valid sample rate would work. + int payload_length_ms = + opus_packet_get_samples_per_frame(payload, 48000) / 48; + if (payload_length_ms < 10) + payload_length_ms = 10; + + int silk_frames; + switch (payload_length_ms) { + case 10: + case 20: + silk_frames = 1; + break; + case 40: + silk_frames = 2; + break; + case 60: + silk_frames = 3; + break; + default: + return 0; // It is actually even an invalid packet. + } + return silk_frames; +} + +// This method is based on Definition of the Opus Audio Codec +// (https://tools.ietf.org/html/rfc6716). Basically, this method is based on +// parsing the LP layer of an Opus packet, particularly the LBRR flag. +int WebRtcOpus_PacketHasFec(const uint8_t* payload, + size_t payload_length_bytes) { + if (payload == NULL || payload_length_bytes == 0) + return 0; + + // In CELT_ONLY mode, packets should not have FEC. + if (payload[0] & 0x80) + return 0; + + int silk_frames = WebRtcOpus_NumSilkFrames(payload); + if (silk_frames == 0) + return 0; // Not valid. + + const int channels = opus_packet_get_nb_channels(payload); + RTC_DCHECK(channels == 1 || channels == 2); + + // Max number of frames in an Opus packet is 48. + opus_int16 frame_sizes[48]; + const unsigned char* frame_data[48]; + + // Parse packet to get the frames. But we only care about the first frame, + // since we can only decode the FEC from the first one. + if (opus_packet_parse(payload, static_cast(payload_length_bytes), + NULL, frame_data, frame_sizes, NULL) < 0) { + return 0; + } + + if (frame_sizes[0] < 1) { + return 0; + } + + // A frame starts with the LP layer. The LP layer begins with two to eight + // header bits.These consist of one VAD bit per SILK frame (up to 3), + // followed by a single flag indicating the presence of LBRR frames. + // For a stereo packet, these first flags correspond to the mid channel, and + // a second set of flags is included for the side channel. Because these are + // the first symbols decoded by the range coder and because they are coded + // as binary values with uniform probability, they can be extracted directly + // from the most significant bits of the first byte of compressed data. + for (int n = 0; n < channels; n++) { + // The LBRR bit for channel 1 is on the (`silk_frames` + 1)-th bit, and + // that of channel 2 is on the |(`silk_frames` + 1) * 2 + 1|-th bit. + if (frame_data[0][0] & (0x80 >> ((n + 1) * (silk_frames + 1) - 1))) + return 1; + } + + return 0; +} + +int WebRtcOpus_PacketHasVoiceActivity(const uint8_t* payload, + size_t payload_length_bytes) { + if (payload == NULL || payload_length_bytes == 0) + return 0; + + // In CELT_ONLY mode we can not determine whether there is VAD. + if (payload[0] & 0x80) + return -1; + + int silk_frames = WebRtcOpus_NumSilkFrames(payload); + if (silk_frames == 0) + return -1; + + const int channels = opus_packet_get_nb_channels(payload); + RTC_DCHECK(channels == 1 || channels == 2); + + // Max number of frames in an Opus packet is 48. + opus_int16 frame_sizes[48]; + const unsigned char* frame_data[48]; + + // Parse packet to get the frames. + int frames = + opus_packet_parse(payload, static_cast(payload_length_bytes), + NULL, frame_data, frame_sizes, NULL); + if (frames < 0) + return -1; + + // Iterate over all Opus frames which may contain multiple SILK frames. + for (int frame = 0; frame < frames; frame++) { + if (frame_sizes[frame] < 1) { + continue; + } + if (frame_data[frame][0] >> (8 - silk_frames)) + return 1; + if (channels == 2 && + (frame_data[frame][0] << (silk_frames + 1)) >> (8 - silk_frames)) + return 1; + } + + return 0; +} diff --git a/third_party/libwebrtc/modules/audio_coding/codecs/opus/opus_interface.h b/third_party/libwebrtc/modules/audio_coding/codecs/opus/opus_interface.h new file mode 100644 index 0000000000..89159ce1c0 --- /dev/null +++ b/third_party/libwebrtc/modules/audio_coding/codecs/opus/opus_interface.h @@ -0,0 +1,547 @@ +/* + * Copyright (c) 2012 The WebRTC project authors. All Rights Reserved. + * + * Use of this source code is governed by a BSD-style license + * that can be found in the LICENSE file in the root of the source + * tree. An additional intellectual property rights grant can be found + * in the file PATENTS. All contributing project authors may + * be found in the AUTHORS file in the root of the source tree. + */ + +#ifndef MODULES_AUDIO_CODING_CODECS_OPUS_OPUS_INTERFACE_H_ +#define MODULES_AUDIO_CODING_CODECS_OPUS_OPUS_INTERFACE_H_ + +#include +#include + +#include "modules/audio_coding/codecs/opus/opus_inst.h" + +#ifdef __cplusplus +extern "C" { +#endif + +// Opaque wrapper types for the codec state. +typedef struct WebRtcOpusEncInst OpusEncInst; +typedef struct WebRtcOpusDecInst OpusDecInst; + +/**************************************************************************** + * WebRtcOpus_EncoderCreate(...) + * + * This function creates an Opus encoder that encodes mono or stereo. + * + * Input: + * - channels : number of channels; 1 or 2. + * - application : 0 - VOIP applications. + * Favor speech intelligibility. + * 1 - Audio applications. + * Favor faithfulness to the original input. + * - sample_rate_hz : sample rate of input audio + * + * Output: + * - inst : a pointer to Encoder context that is created + * if success. + * + * Return value : 0 - Success + * -1 - Error + */ +int16_t WebRtcOpus_EncoderCreate(OpusEncInst** inst, + size_t channels, + int32_t application, + int sample_rate_hz); + +/**************************************************************************** + * WebRtcOpus_MultistreamEncoderCreate(...) + * + * This function creates an Opus encoder with any supported channel count. + * + * Input: + * - channels : number of channels in the input of the encoder. + * - application : 0 - VOIP applications. + * Favor speech intelligibility. + * 1 - Audio applications. + * Favor faithfulness to the original input. + * - streams : number of streams, as described in RFC 7845. + * - coupled_streams : number of coupled streams, as described in + * RFC 7845. + * - channel_mapping : the channel mapping; pointer to array of + * `channel` bytes, as described in RFC 7845. + * + * Output: + * - inst : a pointer to Encoder context that is created + * if success. + * + * Return value : 0 - Success + * -1 - Error + */ +int16_t WebRtcOpus_MultistreamEncoderCreate( + OpusEncInst** inst, + size_t channels, + int32_t application, + size_t streams, + size_t coupled_streams, + const unsigned char* channel_mapping); + +int16_t WebRtcOpus_EncoderFree(OpusEncInst* inst); + +/**************************************************************************** + * WebRtcOpus_Encode(...) + * + * This function encodes audio as a series of Opus frames and inserts + * it into a packet. Input buffer can be any length. + * + * Input: + * - inst : Encoder context + * - audio_in : Input speech data buffer + * - samples : Samples per channel in audio_in + * - length_encoded_buffer : Output buffer size + * + * Output: + * - encoded : Output compressed data buffer + * + * Return value : >=0 - Length (in bytes) of coded data + * -1 - Error + */ +int WebRtcOpus_Encode(OpusEncInst* inst, + const int16_t* audio_in, + size_t samples, + size_t length_encoded_buffer, + uint8_t* encoded); + +/**************************************************************************** + * WebRtcOpus_SetBitRate(...) + * + * This function adjusts the target bitrate of the encoder. + * + * Input: + * - inst : Encoder context + * - rate : New target bitrate + * + * Return value : 0 - Success + * -1 - Error + */ +int16_t WebRtcOpus_SetBitRate(OpusEncInst* inst, int32_t rate); + +/**************************************************************************** + * WebRtcOpus_SetPacketLossRate(...) + * + * This function configures the encoder's expected packet loss percentage. + * + * Input: + * - inst : Encoder context + * - loss_rate : loss percentage in the range 0-100, inclusive. + * Return value : 0 - Success + * -1 - Error + */ +int16_t WebRtcOpus_SetPacketLossRate(OpusEncInst* inst, int32_t loss_rate); + +/**************************************************************************** + * WebRtcOpus_SetMaxPlaybackRate(...) + * + * Configures the maximum playback rate for encoding. Due to hardware + * limitations, the receiver may render audio up to a playback rate. Opus + * encoder can use this information to optimize for network usage and encoding + * complexity. This will affect the audio bandwidth in the coded audio. However, + * the input/output sample rate is not affected. + * + * Input: + * - inst : Encoder context + * - frequency_hz : Maximum playback rate in Hz. + * This parameter can take any value. The relation + * between the value and the Opus internal mode is + * as following: + * frequency_hz <= 8000 narrow band + * 8000 < frequency_hz <= 12000 medium band + * 12000 < frequency_hz <= 16000 wide band + * 16000 < frequency_hz <= 24000 super wide band + * frequency_hz > 24000 full band + * Return value : 0 - Success + * -1 - Error + */ +int16_t WebRtcOpus_SetMaxPlaybackRate(OpusEncInst* inst, int32_t frequency_hz); + +/**************************************************************************** + * WebRtcOpus_GetMaxPlaybackRate(...) + * + * Queries the maximum playback rate for encoding. If different single-stream + * encoders have different maximum playback rates, this function fails. + * + * Input: + * - inst : Encoder context. + * Output: + * - result_hz : The maximum playback rate in Hz. + * Return value : 0 - Success + * -1 - Error + */ +int16_t WebRtcOpus_GetMaxPlaybackRate(OpusEncInst* const inst, + int32_t* result_hz); + +/* TODO(minyue): Check whether an API to check the FEC and the packet loss rate + * is needed. It might not be very useful since there are not many use cases and + * the caller can always maintain the states. */ + +/**************************************************************************** + * WebRtcOpus_EnableFec() + * + * This function enables FEC for encoding. + * + * Input: + * - inst : Encoder context + * + * Return value : 0 - Success + * -1 - Error + */ +int16_t WebRtcOpus_EnableFec(OpusEncInst* inst); + +/**************************************************************************** + * WebRtcOpus_DisableFec() + * + * This function disables FEC for encoding. + * + * Input: + * - inst : Encoder context + * + * Return value : 0 - Success + * -1 - Error + */ +int16_t WebRtcOpus_DisableFec(OpusEncInst* inst); + +/**************************************************************************** + * WebRtcOpus_EnableDtx() + * + * This function enables Opus internal DTX for encoding. + * + * Input: + * - inst : Encoder context + * + * Return value : 0 - Success + * -1 - Error + */ +int16_t WebRtcOpus_EnableDtx(OpusEncInst* inst); + +/**************************************************************************** + * WebRtcOpus_DisableDtx() + * + * This function disables Opus internal DTX for encoding. + * + * Input: + * - inst : Encoder context + * + * Return value : 0 - Success + * -1 - Error + */ +int16_t WebRtcOpus_DisableDtx(OpusEncInst* inst); + +/**************************************************************************** + * WebRtcOpus_GetUseDtx() + * + * This function gets the DTX configuration used for encoding. + * + * Input: + * - inst : Encoder context + * + * Return value : 0 - Encoder does not use DTX. + * 1 - Encoder uses DTX. + * -1 - Error. + */ +int16_t WebRtcOpus_GetUseDtx(OpusEncInst* inst); + +/**************************************************************************** + * WebRtcOpus_EnableCbr() + * + * This function enables CBR for encoding. + * + * Input: + * - inst : Encoder context + * + * Return value : 0 - Success + * -1 - Error + */ +int16_t WebRtcOpus_EnableCbr(OpusEncInst* inst); + +/**************************************************************************** + * WebRtcOpus_DisableCbr() + * + * This function disables CBR for encoding. + * + * Input: + * - inst : Encoder context + * + * Return value : 0 - Success + * -1 - Error + */ +int16_t WebRtcOpus_DisableCbr(OpusEncInst* inst); + +/* + * WebRtcOpus_SetComplexity(...) + * + * This function adjusts the computational complexity. The effect is the same as + * calling the complexity setting of Opus as an Opus encoder related CTL. + * + * Input: + * - inst : Encoder context + * - complexity : New target complexity (0-10, inclusive) + * + * Return value : 0 - Success + * -1 - Error + */ +int16_t WebRtcOpus_SetComplexity(OpusEncInst* inst, int32_t complexity); + +/* + * WebRtcOpus_GetBandwidth(...) + * + * This function returns the current bandwidth. + * + * Input: + * - inst : Encoder context + * + * Return value : Bandwidth - Success + * -1 - Error + */ +int32_t WebRtcOpus_GetBandwidth(OpusEncInst* inst); + +/* + * WebRtcOpus_SetBandwidth(...) + * + * By default Opus decides which bandwidth to encode the signal in depending on + * the the bitrate. This function overrules the previous setting and forces the + * encoder to encode in narrowband/wideband/fullband/etc. + * + * Input: + * - inst : Encoder context + * - bandwidth : New target bandwidth. Valid values are: + * OPUS_BANDWIDTH_NARROWBAND + * OPUS_BANDWIDTH_MEDIUMBAND + * OPUS_BANDWIDTH_WIDEBAND + * OPUS_BANDWIDTH_SUPERWIDEBAND + * OPUS_BANDWIDTH_FULLBAND + * + * Return value : 0 - Success + * -1 - Error + */ +int16_t WebRtcOpus_SetBandwidth(OpusEncInst* inst, int32_t bandwidth); + +/* + * WebRtcOpus_GetInDtx(...) + * + * Gets the DTX state of the encoder. + * + * Input: + * - inst : Encoder context + * + * Return value : -1 - Error. + * 1 - Last encoded frame was comfort noise update during DTX. + * 0 - Last encoded frame was encoded with encoder not in DTX. + */ +int32_t WebRtcOpus_GetInDtx(OpusEncInst* inst); + +/* + * WebRtcOpus_SetForceChannels(...) + * + * If the encoder is initialized as a stereo encoder, Opus will by default + * decide whether to encode in mono or stereo based on the bitrate. This + * function overrules the previous setting, and forces the encoder to encode + * in auto/mono/stereo. + * + * If the Encoder is initialized as a mono encoder, and one tries to force + * stereo, the function will return an error. + * + * Input: + * - inst : Encoder context + * - num_channels : 0 - Not forced + * 1 - Mono + * 2 - Stereo + * + * Return value : 0 - Success + * -1 - Error + */ +int16_t WebRtcOpus_SetForceChannels(OpusEncInst* inst, size_t num_channels); + +int16_t WebRtcOpus_DecoderCreate(OpusDecInst** inst, + size_t channels, + int sample_rate_hz); + +/**************************************************************************** + * WebRtcOpus_MultistreamDecoderCreate(...) + * + * This function creates an Opus decoder with any supported channel count. + * + * Input: + * - channels : number of output channels that the decoder + * will produce. + * - streams : number of encoded streams, as described in + * RFC 7845. + * - coupled_streams : number of coupled streams, as described in + * RFC 7845. + * - channel_mapping : the channel mapping; pointer to array of + * `channel` bytes, as described in RFC 7845. + * + * Output: + * - inst : a pointer to a Decoder context that is created + * if success. + * + * Return value : 0 - Success + * -1 - Error + */ +int16_t WebRtcOpus_MultistreamDecoderCreate( + OpusDecInst** inst, + size_t channels, + size_t streams, + size_t coupled_streams, + const unsigned char* channel_mapping); + +int16_t WebRtcOpus_DecoderFree(OpusDecInst* inst); + +/**************************************************************************** + * WebRtcOpus_DecoderChannels(...) + * + * This function returns the number of channels created for Opus decoder. + */ +size_t WebRtcOpus_DecoderChannels(OpusDecInst* inst); + +/**************************************************************************** + * WebRtcOpus_DecoderInit(...) + * + * This function resets state of the decoder. + * + * Input: + * - inst : Decoder context + */ +void WebRtcOpus_DecoderInit(OpusDecInst* inst); + +/**************************************************************************** + * WebRtcOpus_Decode(...) + * + * This function decodes an Opus packet into one or more audio frames at the + * ACM interface's sampling rate (32 kHz). + * + * Input: + * - inst : Decoder context + * - encoded : Encoded data + * - encoded_bytes : Bytes in encoded vector + * + * Output: + * - decoded : The decoded vector + * - audio_type : 1 normal, 2 CNG (for Opus it should + * always return 1 since we're not using Opus's + * built-in DTX/CNG scheme) + * + * Return value : >0 - Samples per channel in decoded vector + * -1 - Error + */ +int WebRtcOpus_Decode(OpusDecInst* inst, + const uint8_t* encoded, + size_t encoded_bytes, + int16_t* decoded, + int16_t* audio_type); + +/**************************************************************************** + * WebRtcOpus_DecodeFec(...) + * + * This function decodes the FEC data from an Opus packet into one or more audio + * frames at the ACM interface's sampling rate (32 kHz). + * + * Input: + * - inst : Decoder context + * - encoded : Encoded data + * - encoded_bytes : Bytes in encoded vector + * + * Output: + * - decoded : The decoded vector (previous frame) + * + * Return value : >0 - Samples per channel in decoded vector + * 0 - No FEC data in the packet + * -1 - Error + */ +int WebRtcOpus_DecodeFec(OpusDecInst* inst, + const uint8_t* encoded, + size_t encoded_bytes, + int16_t* decoded, + int16_t* audio_type); + +/**************************************************************************** + * WebRtcOpus_DurationEst(...) + * + * This function calculates the duration of an opus packet. + * Input: + * - inst : Decoder context + * - payload : Encoded data pointer + * - payload_length_bytes : Bytes of encoded data + * + * Return value : The duration of the packet, in samples per + * channel. + */ +int WebRtcOpus_DurationEst(OpusDecInst* inst, + const uint8_t* payload, + size_t payload_length_bytes); + +/**************************************************************************** + * WebRtcOpus_PlcDuration(...) + * + * This function calculates the duration of a frame returned by packet loss + * concealment (PLC). + * + * Input: + * - inst : Decoder context + * + * Return value : The duration of a frame returned by PLC, in + * samples per channel. + */ +int WebRtcOpus_PlcDuration(OpusDecInst* inst); + +/* TODO(minyue): Check whether it is needed to add a decoder context to the + * arguments, like WebRtcOpus_DurationEst(...). In fact, the packet itself tells + * the duration. The decoder context in WebRtcOpus_DurationEst(...) is not used. + * So it may be advisable to remove it from WebRtcOpus_DurationEst(...). */ + +/**************************************************************************** + * WebRtcOpus_FecDurationEst(...) + * + * This function calculates the duration of the FEC data within an opus packet. + * Input: + * - payload : Encoded data pointer + * - payload_length_bytes : Bytes of encoded data + * - sample_rate_hz : Sample rate of output audio + * + * Return value : >0 - The duration of the FEC data in the + * packet in samples per channel. + * 0 - No FEC data in the packet. + */ +int WebRtcOpus_FecDurationEst(const uint8_t* payload, + size_t payload_length_bytes, + int sample_rate_hz); + +/**************************************************************************** + * WebRtcOpus_PacketHasFec(...) + * + * This function detects if an opus packet has FEC. + * Input: + * - payload : Encoded data pointer + * - payload_length_bytes : Bytes of encoded data + * + * Return value : 0 - the packet does NOT contain FEC. + * 1 - the packet contains FEC. + */ +int WebRtcOpus_PacketHasFec(const uint8_t* payload, + size_t payload_length_bytes); + +/**************************************************************************** + * WebRtcOpus_PacketHasVoiceActivity(...) + * + * This function returns the SILK VAD information encoded in the opus packet. + * For CELT-only packets that do not have VAD information, it returns -1. + * Input: + * - payload : Encoded data pointer + * - payload_length_bytes : Bytes of encoded data + * + * Return value : 0 - no frame had the VAD flag set. + * 1 - at least one frame had the VAD flag set. + * -1 - VAD status could not be determined. + */ +int WebRtcOpus_PacketHasVoiceActivity(const uint8_t* payload, + size_t payload_length_bytes); + +#ifdef __cplusplus +} // extern "C" +#endif + +#endif // MODULES_AUDIO_CODING_CODECS_OPUS_OPUS_INTERFACE_H_ diff --git a/third_party/libwebrtc/modules/audio_coding/codecs/opus/opus_speed_test.cc b/third_party/libwebrtc/modules/audio_coding/codecs/opus/opus_speed_test.cc new file mode 100644 index 0000000000..4477e8a5f8 --- /dev/null +++ b/third_party/libwebrtc/modules/audio_coding/codecs/opus/opus_speed_test.cc @@ -0,0 +1,147 @@ +/* + * Copyright (c) 2014 The WebRTC project authors. All Rights Reserved. + * + * Use of this source code is governed by a BSD-style license + * that can be found in the LICENSE file in the root of the source + * tree. An additional intellectual property rights grant can be found + * in the file PATENTS. All contributing project authors may + * be found in the AUTHORS file in the root of the source tree. + */ + +#include "modules/audio_coding/codecs/opus/opus_interface.h" +#include "modules/audio_coding/codecs/tools/audio_codec_speed_test.h" + +using ::std::string; + +namespace webrtc { + +static const int kOpusBlockDurationMs = 20; +static const int kOpusSamplingKhz = 48; + +class OpusSpeedTest : public AudioCodecSpeedTest { + protected: + OpusSpeedTest(); + void SetUp() override; + void TearDown() override; + float EncodeABlock(int16_t* in_data, + uint8_t* bit_stream, + size_t max_bytes, + size_t* encoded_bytes) override; + float DecodeABlock(const uint8_t* bit_stream, + size_t encoded_bytes, + int16_t* out_data) override; + WebRtcOpusEncInst* opus_encoder_; + WebRtcOpusDecInst* opus_decoder_; +}; + +OpusSpeedTest::OpusSpeedTest() + : AudioCodecSpeedTest(kOpusBlockDurationMs, + kOpusSamplingKhz, + kOpusSamplingKhz), + opus_encoder_(NULL), + opus_decoder_(NULL) {} + +void OpusSpeedTest::SetUp() { + AudioCodecSpeedTest::SetUp(); + // If channels_ == 1, use Opus VOIP mode, otherwise, audio mode. + int app = channels_ == 1 ? 0 : 1; + /* Create encoder memory. */ + EXPECT_EQ(0, WebRtcOpus_EncoderCreate(&opus_encoder_, channels_, app, 48000)); + EXPECT_EQ(0, WebRtcOpus_DecoderCreate(&opus_decoder_, channels_, 48000)); + /* Set bitrate. */ + EXPECT_EQ(0, WebRtcOpus_SetBitRate(opus_encoder_, bit_rate_)); +} + +void OpusSpeedTest::TearDown() { + AudioCodecSpeedTest::TearDown(); + /* Free memory. */ + EXPECT_EQ(0, WebRtcOpus_EncoderFree(opus_encoder_)); + EXPECT_EQ(0, WebRtcOpus_DecoderFree(opus_decoder_)); +} + +float OpusSpeedTest::EncodeABlock(int16_t* in_data, + uint8_t* bit_stream, + size_t max_bytes, + size_t* encoded_bytes) { + clock_t clocks = clock(); + int value = WebRtcOpus_Encode(opus_encoder_, in_data, input_length_sample_, + max_bytes, bit_stream); + clocks = clock() - clocks; + EXPECT_GT(value, 0); + *encoded_bytes = static_cast(value); + return 1000.0 * clocks / CLOCKS_PER_SEC; +} + +float OpusSpeedTest::DecodeABlock(const uint8_t* bit_stream, + size_t encoded_bytes, + int16_t* out_data) { + int value; + int16_t audio_type; + clock_t clocks = clock(); + value = WebRtcOpus_Decode(opus_decoder_, bit_stream, encoded_bytes, out_data, + &audio_type); + clocks = clock() - clocks; + EXPECT_EQ(output_length_sample_, static_cast(value)); + return 1000.0 * clocks / CLOCKS_PER_SEC; +} + +/* Test audio length in second. */ +constexpr size_t kDurationSec = 400; + +#define ADD_TEST(complexity) \ + TEST_P(OpusSpeedTest, OpusSetComplexityTest##complexity) { \ + /* Set complexity. */ \ + printf("Setting complexity to %d ...\n", complexity); \ + EXPECT_EQ(0, WebRtcOpus_SetComplexity(opus_encoder_, complexity)); \ + EncodeDecode(kDurationSec); \ + } + +ADD_TEST(10) +ADD_TEST(9) +ADD_TEST(8) +ADD_TEST(7) +ADD_TEST(6) +ADD_TEST(5) +ADD_TEST(4) +ADD_TEST(3) +ADD_TEST(2) +ADD_TEST(1) +ADD_TEST(0) + +#define ADD_BANDWIDTH_TEST(bandwidth) \ + TEST_P(OpusSpeedTest, OpusSetBandwidthTest##bandwidth) { \ + /* Set bandwidth. */ \ + printf("Setting bandwidth to %d ...\n", bandwidth); \ + EXPECT_EQ(0, WebRtcOpus_SetBandwidth(opus_encoder_, bandwidth)); \ + EncodeDecode(kDurationSec); \ + } + +ADD_BANDWIDTH_TEST(OPUS_BANDWIDTH_NARROWBAND) +ADD_BANDWIDTH_TEST(OPUS_BANDWIDTH_MEDIUMBAND) +ADD_BANDWIDTH_TEST(OPUS_BANDWIDTH_WIDEBAND) +ADD_BANDWIDTH_TEST(OPUS_BANDWIDTH_SUPERWIDEBAND) +ADD_BANDWIDTH_TEST(OPUS_BANDWIDTH_FULLBAND) + +// List all test cases: (channel, bit rat, filename, extension). +const coding_param param_set[] = { + std::make_tuple(1, + 64000, + string("audio_coding/speech_mono_32_48kHz"), + string("pcm"), + true), + std::make_tuple(1, + 32000, + string("audio_coding/speech_mono_32_48kHz"), + string("pcm"), + true), + std::make_tuple(2, + 64000, + string("audio_coding/music_stereo_48kHz"), + string("pcm"), + true)}; + +INSTANTIATE_TEST_SUITE_P(AllTest, + OpusSpeedTest, + ::testing::ValuesIn(param_set)); + +} // namespace webrtc diff --git a/third_party/libwebrtc/modules/audio_coding/codecs/opus/opus_unittest.cc b/third_party/libwebrtc/modules/audio_coding/codecs/opus/opus_unittest.cc new file mode 100644 index 0000000000..4a9156ad58 --- /dev/null +++ b/third_party/libwebrtc/modules/audio_coding/codecs/opus/opus_unittest.cc @@ -0,0 +1,979 @@ +/* + * Copyright (c) 2013 The WebRTC project authors. All Rights Reserved. + * + * Use of this source code is governed by a BSD-style license + * that can be found in the LICENSE file in the root of the source + * tree. An additional intellectual property rights grant can be found + * in the file PATENTS. All contributing project authors may + * be found in the AUTHORS file in the root of the source tree. + */ + +#include +#include + +#include "modules/audio_coding/codecs/opus/opus_inst.h" +#include "modules/audio_coding/codecs/opus/opus_interface.h" +#include "modules/audio_coding/neteq/tools/audio_loop.h" +#include "rtc_base/checks.h" +#include "rtc_base/numerics/safe_conversions.h" +#include "test/gtest.h" +#include "test/testsupport/file_utils.h" + +namespace webrtc { + +namespace { +// Equivalent to SDP params +// {{"channel_mapping", "0,1,2,3"}, {"coupled_streams", "2"}}. +constexpr unsigned char kQuadChannelMapping[] = {0, 1, 2, 3}; +constexpr int kQuadTotalStreams = 2; +constexpr int kQuadCoupledStreams = 2; + +constexpr unsigned char kStereoChannelMapping[] = {0, 1}; +constexpr int kStereoTotalStreams = 1; +constexpr int kStereoCoupledStreams = 1; + +constexpr unsigned char kMonoChannelMapping[] = {0}; +constexpr int kMonoTotalStreams = 1; +constexpr int kMonoCoupledStreams = 0; + +void CreateSingleOrMultiStreamEncoder(WebRtcOpusEncInst** opus_encoder, + int channels, + int application, + bool use_multistream, + int encoder_sample_rate_hz) { + EXPECT_TRUE(channels == 1 || channels == 2 || use_multistream); + if (use_multistream) { + EXPECT_EQ(encoder_sample_rate_hz, 48000); + if (channels == 1) { + EXPECT_EQ(0, WebRtcOpus_MultistreamEncoderCreate( + opus_encoder, channels, application, kMonoTotalStreams, + kMonoCoupledStreams, kMonoChannelMapping)); + } else if (channels == 2) { + EXPECT_EQ(0, WebRtcOpus_MultistreamEncoderCreate( + opus_encoder, channels, application, kStereoTotalStreams, + kStereoCoupledStreams, kStereoChannelMapping)); + } else if (channels == 4) { + EXPECT_EQ(0, WebRtcOpus_MultistreamEncoderCreate( + opus_encoder, channels, application, kQuadTotalStreams, + kQuadCoupledStreams, kQuadChannelMapping)); + } else { + EXPECT_TRUE(false) << channels; + } + } else { + EXPECT_EQ(0, WebRtcOpus_EncoderCreate(opus_encoder, channels, application, + encoder_sample_rate_hz)); + } +} + +void CreateSingleOrMultiStreamDecoder(WebRtcOpusDecInst** opus_decoder, + int channels, + bool use_multistream, + int decoder_sample_rate_hz) { + EXPECT_TRUE(channels == 1 || channels == 2 || use_multistream); + if (use_multistream) { + EXPECT_EQ(decoder_sample_rate_hz, 48000); + if (channels == 1) { + EXPECT_EQ(0, WebRtcOpus_MultistreamDecoderCreate( + opus_decoder, channels, kMonoTotalStreams, + kMonoCoupledStreams, kMonoChannelMapping)); + } else if (channels == 2) { + EXPECT_EQ(0, WebRtcOpus_MultistreamDecoderCreate( + opus_decoder, channels, kStereoTotalStreams, + kStereoCoupledStreams, kStereoChannelMapping)); + } else if (channels == 4) { + EXPECT_EQ(0, WebRtcOpus_MultistreamDecoderCreate( + opus_decoder, channels, kQuadTotalStreams, + kQuadCoupledStreams, kQuadChannelMapping)); + } else { + EXPECT_TRUE(false) << channels; + } + } else { + EXPECT_EQ(0, WebRtcOpus_DecoderCreate(opus_decoder, channels, + decoder_sample_rate_hz)); + } +} + +int SamplesPerChannel(int sample_rate_hz, int duration_ms) { + const int samples_per_ms = rtc::CheckedDivExact(sample_rate_hz, 1000); + return samples_per_ms * duration_ms; +} + +using test::AudioLoop; +using ::testing::Combine; +using ::testing::TestWithParam; +using ::testing::Values; + +// Maximum number of bytes in output bitstream. +const size_t kMaxBytes = 2000; + +class OpusTest + : public TestWithParam<::testing::tuple> { + protected: + OpusTest() = default; + + void TestDtxEffect(bool dtx, int block_length_ms); + + void TestCbrEffect(bool dtx, int block_length_ms); + + // Prepare `speech_data_` for encoding, read from a hard-coded file. + // After preparation, `speech_data_.GetNextBlock()` returns a pointer to a + // block of `block_length_ms` milliseconds. The data is looped every + // `loop_length_ms` milliseconds. + void PrepareSpeechData(int block_length_ms, int loop_length_ms); + + int EncodeDecode(WebRtcOpusEncInst* encoder, + rtc::ArrayView input_audio, + WebRtcOpusDecInst* decoder, + int16_t* output_audio, + int16_t* audio_type); + + void SetMaxPlaybackRate(WebRtcOpusEncInst* encoder, + opus_int32 expect, + int32_t set); + + void CheckAudioBounded(const int16_t* audio, + size_t samples, + size_t channels, + uint16_t bound) const; + + WebRtcOpusEncInst* opus_encoder_ = nullptr; + WebRtcOpusDecInst* opus_decoder_ = nullptr; + AudioLoop speech_data_; + uint8_t bitstream_[kMaxBytes]; + size_t encoded_bytes_ = 0; + const size_t channels_{std::get<0>(GetParam())}; + const int application_{std::get<1>(GetParam())}; + const bool use_multistream_{std::get<2>(GetParam())}; + const int encoder_sample_rate_hz_{std::get<3>(GetParam())}; + const int decoder_sample_rate_hz_{std::get<4>(GetParam())}; +}; + +} // namespace + +// Singlestream: Try all combinations. +INSTANTIATE_TEST_SUITE_P(Singlestream, + OpusTest, + testing::Combine(testing::Values(1, 2), + testing::Values(0, 1), + testing::Values(false), + testing::Values(16000, 48000), + testing::Values(16000, 48000))); + +// Multistream: Some representative cases (only 48 kHz for now). +INSTANTIATE_TEST_SUITE_P( + Multistream, + OpusTest, + testing::Values(std::make_tuple(1, 0, true, 48000, 48000), + std::make_tuple(2, 1, true, 48000, 48000), + std::make_tuple(4, 0, true, 48000, 48000), + std::make_tuple(4, 1, true, 48000, 48000))); + +void OpusTest::PrepareSpeechData(int block_length_ms, int loop_length_ms) { + std::map channel_to_basename = { + {1, "audio_coding/testfile32kHz"}, + {2, "audio_coding/teststereo32kHz"}, + {4, "audio_coding/speech_4_channels_48k_one_second"}}; + std::map channel_to_suffix = { + {1, "pcm"}, {2, "pcm"}, {4, "wav"}}; + const std::string file_name = webrtc::test::ResourcePath( + channel_to_basename[channels_], channel_to_suffix[channels_]); + if (loop_length_ms < block_length_ms) { + loop_length_ms = block_length_ms; + } + const int sample_rate_khz = + rtc::CheckedDivExact(encoder_sample_rate_hz_, 1000); + EXPECT_TRUE(speech_data_.Init(file_name, + loop_length_ms * sample_rate_khz * channels_, + block_length_ms * sample_rate_khz * channels_)); +} + +void OpusTest::SetMaxPlaybackRate(WebRtcOpusEncInst* encoder, + opus_int32 expect, + int32_t set) { + opus_int32 bandwidth; + EXPECT_EQ(0, WebRtcOpus_SetMaxPlaybackRate(opus_encoder_, set)); + EXPECT_EQ(0, WebRtcOpus_GetMaxPlaybackRate(opus_encoder_, &bandwidth)); + EXPECT_EQ(expect, bandwidth); +} + +void OpusTest::CheckAudioBounded(const int16_t* audio, + size_t samples, + size_t channels, + uint16_t bound) const { + for (size_t i = 0; i < samples; ++i) { + for (size_t c = 0; c < channels; ++c) { + ASSERT_GE(audio[i * channels + c], -bound); + ASSERT_LE(audio[i * channels + c], bound); + } + } +} + +int OpusTest::EncodeDecode(WebRtcOpusEncInst* encoder, + rtc::ArrayView input_audio, + WebRtcOpusDecInst* decoder, + int16_t* output_audio, + int16_t* audio_type) { + const int input_samples_per_channel = + rtc::CheckedDivExact(input_audio.size(), channels_); + int encoded_bytes_int = + WebRtcOpus_Encode(encoder, input_audio.data(), input_samples_per_channel, + kMaxBytes, bitstream_); + EXPECT_GE(encoded_bytes_int, 0); + encoded_bytes_ = static_cast(encoded_bytes_int); + if (encoded_bytes_ != 0) { + int est_len = WebRtcOpus_DurationEst(decoder, bitstream_, encoded_bytes_); + int act_len = WebRtcOpus_Decode(decoder, bitstream_, encoded_bytes_, + output_audio, audio_type); + EXPECT_EQ(est_len, act_len); + return act_len; + } else { + int total_dtx_len = 0; + const int output_samples_per_channel = input_samples_per_channel * + decoder_sample_rate_hz_ / + encoder_sample_rate_hz_; + while (total_dtx_len < output_samples_per_channel) { + int est_len = WebRtcOpus_DurationEst(decoder, NULL, 0); + int act_len = WebRtcOpus_Decode(decoder, NULL, 0, + &output_audio[total_dtx_len * channels_], + audio_type); + EXPECT_EQ(est_len, act_len); + total_dtx_len += act_len; + } + return total_dtx_len; + } +} + +// Test if encoder/decoder can enter DTX mode properly and do not enter DTX when +// they should not. This test is signal dependent. +void OpusTest::TestDtxEffect(bool dtx, int block_length_ms) { + PrepareSpeechData(block_length_ms, 2000); + const size_t input_samples = + rtc::CheckedDivExact(encoder_sample_rate_hz_, 1000) * block_length_ms; + const size_t output_samples = + rtc::CheckedDivExact(decoder_sample_rate_hz_, 1000) * block_length_ms; + + // Create encoder memory. + CreateSingleOrMultiStreamEncoder(&opus_encoder_, channels_, application_, + use_multistream_, encoder_sample_rate_hz_); + CreateSingleOrMultiStreamDecoder(&opus_decoder_, channels_, use_multistream_, + decoder_sample_rate_hz_); + + // Set bitrate. + EXPECT_EQ( + 0, WebRtcOpus_SetBitRate(opus_encoder_, channels_ == 1 ? 32000 : 64000)); + + // Set input audio as silence. + std::vector silence(input_samples * channels_, 0); + + // Setting DTX. + EXPECT_EQ(0, dtx ? WebRtcOpus_EnableDtx(opus_encoder_) + : WebRtcOpus_DisableDtx(opus_encoder_)); + + int16_t audio_type; + int16_t* output_data_decode = new int16_t[output_samples * channels_]; + + for (int i = 0; i < 100; ++i) { + EXPECT_EQ(output_samples, + static_cast(EncodeDecode( + opus_encoder_, speech_data_.GetNextBlock(), opus_decoder_, + output_data_decode, &audio_type))); + // If not DTX, it should never enter DTX mode. If DTX, we do not care since + // whether it enters DTX depends on the signal type. + if (!dtx) { + EXPECT_GT(encoded_bytes_, 1U); + EXPECT_EQ(0, opus_encoder_->in_dtx_mode); + EXPECT_EQ(0, opus_decoder_->in_dtx_mode); + EXPECT_EQ(0, audio_type); // Speech. + } + } + + // We input some silent segments. In DTX mode, the encoder will stop sending. + // However, DTX may happen after a while. + for (int i = 0; i < 30; ++i) { + EXPECT_EQ(output_samples, static_cast(EncodeDecode( + opus_encoder_, silence, opus_decoder_, + output_data_decode, &audio_type))); + if (!dtx) { + EXPECT_GT(encoded_bytes_, 1U); + EXPECT_EQ(0, opus_encoder_->in_dtx_mode); + EXPECT_EQ(0, opus_decoder_->in_dtx_mode); + EXPECT_EQ(0, audio_type); // Speech. + } else if (encoded_bytes_ == 1) { + EXPECT_EQ(1, opus_encoder_->in_dtx_mode); + EXPECT_EQ(1, opus_decoder_->in_dtx_mode); + EXPECT_EQ(2, audio_type); // Comfort noise. + break; + } + } + + // When Opus is in DTX, it wakes up in a regular basis. It sends two packets, + // one with an arbitrary size and the other of 1-byte, then stops sending for + // a certain number of frames. + + // `max_dtx_frames` is the maximum number of frames Opus can stay in DTX. + // TODO(kwiberg): Why does this number depend on the encoding sample rate? + const int max_dtx_frames = + (encoder_sample_rate_hz_ == 16000 ? 800 : 400) / block_length_ms + 1; + + // We run `kRunTimeMs` milliseconds of pure silence. + const int kRunTimeMs = 4500; + + // We check that, after a `kCheckTimeMs` milliseconds (given that the CNG in + // Opus needs time to adapt), the absolute values of DTX decoded signal are + // bounded by `kOutputValueBound`. + const int kCheckTimeMs = 4000; + +#if defined(OPUS_FIXED_POINT) + // Fixed-point Opus generates a random (comfort) noise, which has a less + // predictable value bound than its floating-point Opus. This value depends on + // input signal, and the time window for checking the output values (between + // `kCheckTimeMs` and `kRunTimeMs`). + const uint16_t kOutputValueBound = 30; + +#else + const uint16_t kOutputValueBound = 2; +#endif + + int time = 0; + while (time < kRunTimeMs) { + // DTX mode is maintained for maximum `max_dtx_frames` frames. + int i = 0; + for (; i < max_dtx_frames; ++i) { + time += block_length_ms; + EXPECT_EQ(output_samples, static_cast(EncodeDecode( + opus_encoder_, silence, opus_decoder_, + output_data_decode, &audio_type))); + if (dtx) { + if (encoded_bytes_ > 1) + break; + EXPECT_EQ(0U, encoded_bytes_) // Send 0 byte. + << "Opus should have entered DTX mode."; + EXPECT_EQ(1, opus_encoder_->in_dtx_mode); + EXPECT_EQ(1, opus_decoder_->in_dtx_mode); + EXPECT_EQ(2, audio_type); // Comfort noise. + if (time >= kCheckTimeMs) { + CheckAudioBounded(output_data_decode, output_samples, channels_, + kOutputValueBound); + } + } else { + EXPECT_GT(encoded_bytes_, 1U); + EXPECT_EQ(0, opus_encoder_->in_dtx_mode); + EXPECT_EQ(0, opus_decoder_->in_dtx_mode); + EXPECT_EQ(0, audio_type); // Speech. + } + } + + if (dtx) { + // With DTX, Opus must stop transmission for some time. + EXPECT_GT(i, 1); + } + + // We expect a normal payload. + EXPECT_EQ(0, opus_encoder_->in_dtx_mode); + EXPECT_EQ(0, opus_decoder_->in_dtx_mode); + EXPECT_EQ(0, audio_type); // Speech. + + // Enters DTX again immediately. + time += block_length_ms; + EXPECT_EQ(output_samples, static_cast(EncodeDecode( + opus_encoder_, silence, opus_decoder_, + output_data_decode, &audio_type))); + if (dtx) { + EXPECT_EQ(1U, encoded_bytes_); // Send 1 byte. + EXPECT_EQ(1, opus_encoder_->in_dtx_mode); + EXPECT_EQ(1, opus_decoder_->in_dtx_mode); + EXPECT_EQ(2, audio_type); // Comfort noise. + if (time >= kCheckTimeMs) { + CheckAudioBounded(output_data_decode, output_samples, channels_, + kOutputValueBound); + } + } else { + EXPECT_GT(encoded_bytes_, 1U); + EXPECT_EQ(0, opus_encoder_->in_dtx_mode); + EXPECT_EQ(0, opus_decoder_->in_dtx_mode); + EXPECT_EQ(0, audio_type); // Speech. + } + } + + silence[0] = 10000; + if (dtx) { + // Verify that encoder/decoder can jump out from DTX mode. + EXPECT_EQ(output_samples, static_cast(EncodeDecode( + opus_encoder_, silence, opus_decoder_, + output_data_decode, &audio_type))); + EXPECT_GT(encoded_bytes_, 1U); + EXPECT_EQ(0, opus_encoder_->in_dtx_mode); + EXPECT_EQ(0, opus_decoder_->in_dtx_mode); + EXPECT_EQ(0, audio_type); // Speech. + } + + // Free memory. + delete[] output_data_decode; + EXPECT_EQ(0, WebRtcOpus_EncoderFree(opus_encoder_)); + EXPECT_EQ(0, WebRtcOpus_DecoderFree(opus_decoder_)); +} + +// Test if CBR does what we expect. +void OpusTest::TestCbrEffect(bool cbr, int block_length_ms) { + PrepareSpeechData(block_length_ms, 2000); + const size_t output_samples = + rtc::CheckedDivExact(decoder_sample_rate_hz_, 1000) * block_length_ms; + + int32_t max_pkt_size_diff = 0; + int32_t prev_pkt_size = 0; + + // Create encoder memory. + CreateSingleOrMultiStreamEncoder(&opus_encoder_, channels_, application_, + use_multistream_, encoder_sample_rate_hz_); + CreateSingleOrMultiStreamDecoder(&opus_decoder_, channels_, use_multistream_, + decoder_sample_rate_hz_); + + // Set bitrate. + EXPECT_EQ( + 0, WebRtcOpus_SetBitRate(opus_encoder_, channels_ == 1 ? 32000 : 64000)); + + // Setting CBR. + EXPECT_EQ(0, cbr ? WebRtcOpus_EnableCbr(opus_encoder_) + : WebRtcOpus_DisableCbr(opus_encoder_)); + + int16_t audio_type; + std::vector audio_out(output_samples * channels_); + for (int i = 0; i < 100; ++i) { + EXPECT_EQ(output_samples, + static_cast( + EncodeDecode(opus_encoder_, speech_data_.GetNextBlock(), + opus_decoder_, audio_out.data(), &audio_type))); + + if (prev_pkt_size > 0) { + int32_t diff = std::abs((int32_t)encoded_bytes_ - prev_pkt_size); + max_pkt_size_diff = std::max(max_pkt_size_diff, diff); + } + prev_pkt_size = rtc::checked_cast(encoded_bytes_); + } + + if (cbr) { + EXPECT_EQ(max_pkt_size_diff, 0); + } else { + EXPECT_GT(max_pkt_size_diff, 0); + } + + // Free memory. + EXPECT_EQ(0, WebRtcOpus_EncoderFree(opus_encoder_)); + EXPECT_EQ(0, WebRtcOpus_DecoderFree(opus_decoder_)); +} + +// Test failing Create. +TEST(OpusTest, OpusCreateFail) { + WebRtcOpusEncInst* opus_encoder; + WebRtcOpusDecInst* opus_decoder; + + // Test to see that an invalid pointer is caught. + EXPECT_EQ(-1, WebRtcOpus_EncoderCreate(NULL, 1, 0, 48000)); + // Invalid channel number. + EXPECT_EQ(-1, WebRtcOpus_EncoderCreate(&opus_encoder, 257, 0, 48000)); + // Invalid applciation mode. + EXPECT_EQ(-1, WebRtcOpus_EncoderCreate(&opus_encoder, 1, 2, 48000)); + // Invalid sample rate. + EXPECT_EQ(-1, WebRtcOpus_EncoderCreate(&opus_encoder, 1, 0, 12345)); + + EXPECT_EQ(-1, WebRtcOpus_DecoderCreate(NULL, 1, 48000)); + // Invalid channel number. + EXPECT_EQ(-1, WebRtcOpus_DecoderCreate(&opus_decoder, 257, 48000)); + // Invalid sample rate. + EXPECT_EQ(-1, WebRtcOpus_DecoderCreate(&opus_decoder, 1, 12345)); +} + +// Test failing Free. +TEST(OpusTest, OpusFreeFail) { + // Test to see that an invalid pointer is caught. + EXPECT_EQ(-1, WebRtcOpus_EncoderFree(NULL)); + EXPECT_EQ(-1, WebRtcOpus_DecoderFree(NULL)); +} + +// Test normal Create and Free. +TEST_P(OpusTest, OpusCreateFree) { + CreateSingleOrMultiStreamEncoder(&opus_encoder_, channels_, application_, + use_multistream_, encoder_sample_rate_hz_); + CreateSingleOrMultiStreamDecoder(&opus_decoder_, channels_, use_multistream_, + decoder_sample_rate_hz_); + EXPECT_TRUE(opus_encoder_ != NULL); + EXPECT_TRUE(opus_decoder_ != NULL); + // Free encoder and decoder memory. + EXPECT_EQ(0, WebRtcOpus_EncoderFree(opus_encoder_)); + EXPECT_EQ(0, WebRtcOpus_DecoderFree(opus_decoder_)); +} + +#define ENCODER_CTL(inst, vargs) \ + inst->encoder \ + ? opus_encoder_ctl(inst->encoder, vargs) \ + : opus_multistream_encoder_ctl(inst->multistream_encoder, vargs) + +TEST_P(OpusTest, OpusEncodeDecode) { + PrepareSpeechData(20, 20); + + // Create encoder memory. + CreateSingleOrMultiStreamEncoder(&opus_encoder_, channels_, application_, + use_multistream_, encoder_sample_rate_hz_); + CreateSingleOrMultiStreamDecoder(&opus_decoder_, channels_, use_multistream_, + decoder_sample_rate_hz_); + + // Set bitrate. + EXPECT_EQ( + 0, WebRtcOpus_SetBitRate(opus_encoder_, channels_ == 1 ? 32000 : 64000)); + + // Check number of channels for decoder. + EXPECT_EQ(channels_, WebRtcOpus_DecoderChannels(opus_decoder_)); + + // Check application mode. + opus_int32 app; + ENCODER_CTL(opus_encoder_, OPUS_GET_APPLICATION(&app)); + EXPECT_EQ(application_ == 0 ? OPUS_APPLICATION_VOIP : OPUS_APPLICATION_AUDIO, + app); + + // Encode & decode. + int16_t audio_type; + const int decode_samples_per_channel = + SamplesPerChannel(decoder_sample_rate_hz_, /*ms=*/20); + int16_t* output_data_decode = + new int16_t[decode_samples_per_channel * channels_]; + EXPECT_EQ(decode_samples_per_channel, + EncodeDecode(opus_encoder_, speech_data_.GetNextBlock(), + opus_decoder_, output_data_decode, &audio_type)); + + // Free memory. + delete[] output_data_decode; + EXPECT_EQ(0, WebRtcOpus_EncoderFree(opus_encoder_)); + EXPECT_EQ(0, WebRtcOpus_DecoderFree(opus_decoder_)); +} + +TEST_P(OpusTest, OpusSetBitRate) { + // Test without creating encoder memory. + EXPECT_EQ(-1, WebRtcOpus_SetBitRate(opus_encoder_, 60000)); + + // Create encoder memory, try with different bitrates. + CreateSingleOrMultiStreamEncoder(&opus_encoder_, channels_, application_, + use_multistream_, encoder_sample_rate_hz_); + EXPECT_EQ(0, WebRtcOpus_SetBitRate(opus_encoder_, 30000)); + EXPECT_EQ(0, WebRtcOpus_SetBitRate(opus_encoder_, 60000)); + EXPECT_EQ(0, WebRtcOpus_SetBitRate(opus_encoder_, 300000)); + EXPECT_EQ(0, WebRtcOpus_SetBitRate(opus_encoder_, 600000)); + + // Free memory. + EXPECT_EQ(0, WebRtcOpus_EncoderFree(opus_encoder_)); +} + +TEST_P(OpusTest, OpusSetComplexity) { + // Test without creating encoder memory. + EXPECT_EQ(-1, WebRtcOpus_SetComplexity(opus_encoder_, 9)); + + // Create encoder memory, try with different complexities. + CreateSingleOrMultiStreamEncoder(&opus_encoder_, channels_, application_, + use_multistream_, encoder_sample_rate_hz_); + + EXPECT_EQ(0, WebRtcOpus_SetComplexity(opus_encoder_, 0)); + EXPECT_EQ(0, WebRtcOpus_SetComplexity(opus_encoder_, 10)); + EXPECT_EQ(-1, WebRtcOpus_SetComplexity(opus_encoder_, 11)); + + // Free memory. + EXPECT_EQ(0, WebRtcOpus_EncoderFree(opus_encoder_)); +} + +TEST_P(OpusTest, OpusSetBandwidth) { + if (channels_ > 2) { + // TODO(webrtc:10217): investigate why multi-stream Opus reports + // narrowband when it's configured with FULLBAND. + return; + } + PrepareSpeechData(20, 20); + + int16_t audio_type; + const int decode_samples_per_channel = + SamplesPerChannel(decoder_sample_rate_hz_, /*ms=*/20); + std::unique_ptr output_data_decode( + new int16_t[decode_samples_per_channel * channels_]()); + + // Test without creating encoder memory. + EXPECT_EQ(-1, + WebRtcOpus_SetBandwidth(opus_encoder_, OPUS_BANDWIDTH_NARROWBAND)); + EXPECT_EQ(-1, WebRtcOpus_GetBandwidth(opus_encoder_)); + + // Create encoder memory, try with different bandwidths. + CreateSingleOrMultiStreamEncoder(&opus_encoder_, channels_, application_, + use_multistream_, encoder_sample_rate_hz_); + CreateSingleOrMultiStreamDecoder(&opus_decoder_, channels_, use_multistream_, + decoder_sample_rate_hz_); + + EXPECT_EQ(-1, WebRtcOpus_SetBandwidth(opus_encoder_, + OPUS_BANDWIDTH_NARROWBAND - 1)); + EXPECT_EQ(0, + WebRtcOpus_SetBandwidth(opus_encoder_, OPUS_BANDWIDTH_NARROWBAND)); + EncodeDecode(opus_encoder_, speech_data_.GetNextBlock(), opus_decoder_, + output_data_decode.get(), &audio_type); + EXPECT_EQ(OPUS_BANDWIDTH_NARROWBAND, WebRtcOpus_GetBandwidth(opus_encoder_)); + EXPECT_EQ(0, WebRtcOpus_SetBandwidth(opus_encoder_, OPUS_BANDWIDTH_FULLBAND)); + EncodeDecode(opus_encoder_, speech_data_.GetNextBlock(), opus_decoder_, + output_data_decode.get(), &audio_type); + EXPECT_EQ(encoder_sample_rate_hz_ == 16000 ? OPUS_BANDWIDTH_WIDEBAND + : OPUS_BANDWIDTH_FULLBAND, + WebRtcOpus_GetBandwidth(opus_encoder_)); + EXPECT_EQ( + -1, WebRtcOpus_SetBandwidth(opus_encoder_, OPUS_BANDWIDTH_FULLBAND + 1)); + EncodeDecode(opus_encoder_, speech_data_.GetNextBlock(), opus_decoder_, + output_data_decode.get(), &audio_type); + EXPECT_EQ(encoder_sample_rate_hz_ == 16000 ? OPUS_BANDWIDTH_WIDEBAND + : OPUS_BANDWIDTH_FULLBAND, + WebRtcOpus_GetBandwidth(opus_encoder_)); + + // Free memory. + EXPECT_EQ(0, WebRtcOpus_EncoderFree(opus_encoder_)); + EXPECT_EQ(0, WebRtcOpus_DecoderFree(opus_decoder_)); +} + +TEST_P(OpusTest, OpusForceChannels) { + // Test without creating encoder memory. + EXPECT_EQ(-1, WebRtcOpus_SetForceChannels(opus_encoder_, 1)); + + CreateSingleOrMultiStreamEncoder(&opus_encoder_, channels_, application_, + use_multistream_, encoder_sample_rate_hz_); + ASSERT_NE(nullptr, opus_encoder_); + + if (channels_ >= 2) { + EXPECT_EQ(-1, WebRtcOpus_SetForceChannels(opus_encoder_, 3)); + EXPECT_EQ(0, WebRtcOpus_SetForceChannels(opus_encoder_, 2)); + EXPECT_EQ(0, WebRtcOpus_SetForceChannels(opus_encoder_, 1)); + EXPECT_EQ(0, WebRtcOpus_SetForceChannels(opus_encoder_, 0)); + } else { + EXPECT_EQ(-1, WebRtcOpus_SetForceChannels(opus_encoder_, 2)); + EXPECT_EQ(0, WebRtcOpus_SetForceChannels(opus_encoder_, 1)); + EXPECT_EQ(0, WebRtcOpus_SetForceChannels(opus_encoder_, 0)); + } + + EXPECT_EQ(0, WebRtcOpus_EncoderFree(opus_encoder_)); +} + +// Encode and decode one frame, initialize the decoder and +// decode once more. +TEST_P(OpusTest, OpusDecodeInit) { + PrepareSpeechData(20, 20); + + // Create encoder memory. + CreateSingleOrMultiStreamEncoder(&opus_encoder_, channels_, application_, + use_multistream_, encoder_sample_rate_hz_); + CreateSingleOrMultiStreamDecoder(&opus_decoder_, channels_, use_multistream_, + decoder_sample_rate_hz_); + + // Encode & decode. + int16_t audio_type; + const int decode_samples_per_channel = + SamplesPerChannel(decoder_sample_rate_hz_, /*ms=*/20); + int16_t* output_data_decode = + new int16_t[decode_samples_per_channel * channels_]; + EXPECT_EQ(decode_samples_per_channel, + EncodeDecode(opus_encoder_, speech_data_.GetNextBlock(), + opus_decoder_, output_data_decode, &audio_type)); + + WebRtcOpus_DecoderInit(opus_decoder_); + + EXPECT_EQ(decode_samples_per_channel, + WebRtcOpus_Decode(opus_decoder_, bitstream_, encoded_bytes_, + output_data_decode, &audio_type)); + + // Free memory. + delete[] output_data_decode; + EXPECT_EQ(0, WebRtcOpus_EncoderFree(opus_encoder_)); + EXPECT_EQ(0, WebRtcOpus_DecoderFree(opus_decoder_)); +} + +TEST_P(OpusTest, OpusEnableDisableFec) { + // Test without creating encoder memory. + EXPECT_EQ(-1, WebRtcOpus_EnableFec(opus_encoder_)); + EXPECT_EQ(-1, WebRtcOpus_DisableFec(opus_encoder_)); + + // Create encoder memory. + CreateSingleOrMultiStreamEncoder(&opus_encoder_, channels_, application_, + use_multistream_, encoder_sample_rate_hz_); + + EXPECT_EQ(0, WebRtcOpus_EnableFec(opus_encoder_)); + EXPECT_EQ(0, WebRtcOpus_DisableFec(opus_encoder_)); + + // Free memory. + EXPECT_EQ(0, WebRtcOpus_EncoderFree(opus_encoder_)); +} + +TEST_P(OpusTest, OpusEnableDisableDtx) { + // Test without creating encoder memory. + EXPECT_EQ(-1, WebRtcOpus_EnableDtx(opus_encoder_)); + EXPECT_EQ(-1, WebRtcOpus_DisableDtx(opus_encoder_)); + + // Create encoder memory. + CreateSingleOrMultiStreamEncoder(&opus_encoder_, channels_, application_, + use_multistream_, encoder_sample_rate_hz_); + + opus_int32 dtx; + + // DTX is off by default. + ENCODER_CTL(opus_encoder_, OPUS_GET_DTX(&dtx)); + EXPECT_EQ(0, dtx); + + // Test to enable DTX. + EXPECT_EQ(0, WebRtcOpus_EnableDtx(opus_encoder_)); + ENCODER_CTL(opus_encoder_, OPUS_GET_DTX(&dtx)); + EXPECT_EQ(1, dtx); + + // Test to disable DTX. + EXPECT_EQ(0, WebRtcOpus_DisableDtx(opus_encoder_)); + ENCODER_CTL(opus_encoder_, OPUS_GET_DTX(&dtx)); + EXPECT_EQ(0, dtx); + + // Free memory. + EXPECT_EQ(0, WebRtcOpus_EncoderFree(opus_encoder_)); +} + +TEST_P(OpusTest, OpusDtxOff) { + TestDtxEffect(false, 10); + TestDtxEffect(false, 20); + TestDtxEffect(false, 40); +} + +TEST_P(OpusTest, OpusDtxOn) { + if (channels_ > 2 || application_ != 0) { + // DTX does not work with OPUS_APPLICATION_AUDIO at low complexity settings. + // TODO(webrtc:10218): adapt the test to the sizes and order of multi-stream + // DTX packets. + return; + } + TestDtxEffect(true, 10); + TestDtxEffect(true, 20); + TestDtxEffect(true, 40); +} + +TEST_P(OpusTest, OpusCbrOff) { + TestCbrEffect(false, 10); + TestCbrEffect(false, 20); + TestCbrEffect(false, 40); +} + +TEST_P(OpusTest, OpusCbrOn) { + TestCbrEffect(true, 10); + TestCbrEffect(true, 20); + TestCbrEffect(true, 40); +} + +TEST_P(OpusTest, OpusSetPacketLossRate) { + // Test without creating encoder memory. + EXPECT_EQ(-1, WebRtcOpus_SetPacketLossRate(opus_encoder_, 50)); + + // Create encoder memory. + CreateSingleOrMultiStreamEncoder(&opus_encoder_, channels_, application_, + use_multistream_, encoder_sample_rate_hz_); + + EXPECT_EQ(0, WebRtcOpus_SetPacketLossRate(opus_encoder_, 50)); + EXPECT_EQ(-1, WebRtcOpus_SetPacketLossRate(opus_encoder_, -1)); + EXPECT_EQ(-1, WebRtcOpus_SetPacketLossRate(opus_encoder_, 101)); + + // Free memory. + EXPECT_EQ(0, WebRtcOpus_EncoderFree(opus_encoder_)); +} + +TEST_P(OpusTest, OpusSetMaxPlaybackRate) { + // Test without creating encoder memory. + EXPECT_EQ(-1, WebRtcOpus_SetMaxPlaybackRate(opus_encoder_, 20000)); + + // Create encoder memory. + CreateSingleOrMultiStreamEncoder(&opus_encoder_, channels_, application_, + use_multistream_, encoder_sample_rate_hz_); + + SetMaxPlaybackRate(opus_encoder_, OPUS_BANDWIDTH_FULLBAND, 48000); + SetMaxPlaybackRate(opus_encoder_, OPUS_BANDWIDTH_FULLBAND, 24001); + SetMaxPlaybackRate(opus_encoder_, OPUS_BANDWIDTH_SUPERWIDEBAND, 24000); + SetMaxPlaybackRate(opus_encoder_, OPUS_BANDWIDTH_SUPERWIDEBAND, 16001); + SetMaxPlaybackRate(opus_encoder_, OPUS_BANDWIDTH_WIDEBAND, 16000); + SetMaxPlaybackRate(opus_encoder_, OPUS_BANDWIDTH_WIDEBAND, 12001); + SetMaxPlaybackRate(opus_encoder_, OPUS_BANDWIDTH_MEDIUMBAND, 12000); + SetMaxPlaybackRate(opus_encoder_, OPUS_BANDWIDTH_MEDIUMBAND, 8001); + SetMaxPlaybackRate(opus_encoder_, OPUS_BANDWIDTH_NARROWBAND, 8000); + SetMaxPlaybackRate(opus_encoder_, OPUS_BANDWIDTH_NARROWBAND, 4000); + + // Free memory. + EXPECT_EQ(0, WebRtcOpus_EncoderFree(opus_encoder_)); +} + +// Test PLC. +TEST_P(OpusTest, OpusDecodePlc) { + PrepareSpeechData(20, 20); + + // Create encoder memory. + CreateSingleOrMultiStreamEncoder(&opus_encoder_, channels_, application_, + use_multistream_, encoder_sample_rate_hz_); + CreateSingleOrMultiStreamDecoder(&opus_decoder_, channels_, use_multistream_, + decoder_sample_rate_hz_); + + // Set bitrate. + EXPECT_EQ( + 0, WebRtcOpus_SetBitRate(opus_encoder_, channels_ == 1 ? 32000 : 64000)); + + // Check number of channels for decoder. + EXPECT_EQ(channels_, WebRtcOpus_DecoderChannels(opus_decoder_)); + + // Encode & decode. + int16_t audio_type; + const int decode_samples_per_channel = + SamplesPerChannel(decoder_sample_rate_hz_, /*ms=*/20); + int16_t* output_data_decode = + new int16_t[decode_samples_per_channel * channels_]; + EXPECT_EQ(decode_samples_per_channel, + EncodeDecode(opus_encoder_, speech_data_.GetNextBlock(), + opus_decoder_, output_data_decode, &audio_type)); + + // Call decoder PLC. + constexpr int kPlcDurationMs = 10; + const int plc_samples = decoder_sample_rate_hz_ * kPlcDurationMs / 1000; + int16_t* plc_buffer = new int16_t[plc_samples * channels_]; + EXPECT_EQ(plc_samples, + WebRtcOpus_Decode(opus_decoder_, NULL, 0, plc_buffer, &audio_type)); + + // Free memory. + delete[] plc_buffer; + delete[] output_data_decode; + EXPECT_EQ(0, WebRtcOpus_EncoderFree(opus_encoder_)); + EXPECT_EQ(0, WebRtcOpus_DecoderFree(opus_decoder_)); +} + +// Duration estimation. +TEST_P(OpusTest, OpusDurationEstimation) { + PrepareSpeechData(20, 20); + + // Create. + CreateSingleOrMultiStreamEncoder(&opus_encoder_, channels_, application_, + use_multistream_, encoder_sample_rate_hz_); + CreateSingleOrMultiStreamDecoder(&opus_decoder_, channels_, use_multistream_, + decoder_sample_rate_hz_); + + // 10 ms. We use only first 10 ms of a 20 ms block. + auto speech_block = speech_data_.GetNextBlock(); + int encoded_bytes_int = WebRtcOpus_Encode( + opus_encoder_, speech_block.data(), + rtc::CheckedDivExact(speech_block.size(), 2 * channels_), kMaxBytes, + bitstream_); + EXPECT_GE(encoded_bytes_int, 0); + EXPECT_EQ(SamplesPerChannel(decoder_sample_rate_hz_, /*ms=*/10), + WebRtcOpus_DurationEst(opus_decoder_, bitstream_, + static_cast(encoded_bytes_int))); + + // 20 ms + speech_block = speech_data_.GetNextBlock(); + encoded_bytes_int = + WebRtcOpus_Encode(opus_encoder_, speech_block.data(), + rtc::CheckedDivExact(speech_block.size(), channels_), + kMaxBytes, bitstream_); + EXPECT_GE(encoded_bytes_int, 0); + EXPECT_EQ(SamplesPerChannel(decoder_sample_rate_hz_, /*ms=*/20), + WebRtcOpus_DurationEst(opus_decoder_, bitstream_, + static_cast(encoded_bytes_int))); + + // Free memory. + EXPECT_EQ(0, WebRtcOpus_EncoderFree(opus_encoder_)); + EXPECT_EQ(0, WebRtcOpus_DecoderFree(opus_decoder_)); +} + +TEST_P(OpusTest, OpusDecodeRepacketized) { + if (channels_ > 2) { + // As per the Opus documentation + // https://mf4.xiph.org/jenkins/view/opus/job/opus/ws/doc/html/group__opus__repacketizer.html#details, + // multiple streams are not supported. + return; + } + constexpr size_t kPackets = 6; + + PrepareSpeechData(20, 20 * kPackets); + + // Create encoder memory. + CreateSingleOrMultiStreamEncoder(&opus_encoder_, channels_, application_, + use_multistream_, encoder_sample_rate_hz_); + ASSERT_NE(nullptr, opus_encoder_); + CreateSingleOrMultiStreamDecoder(&opus_decoder_, channels_, use_multistream_, + decoder_sample_rate_hz_); + ASSERT_NE(nullptr, opus_decoder_); + + // Set bitrate. + EXPECT_EQ( + 0, WebRtcOpus_SetBitRate(opus_encoder_, channels_ == 1 ? 32000 : 64000)); + + // Check number of channels for decoder. + EXPECT_EQ(channels_, WebRtcOpus_DecoderChannels(opus_decoder_)); + + // Encode & decode. + int16_t audio_type; + const int decode_samples_per_channel = + SamplesPerChannel(decoder_sample_rate_hz_, /*ms=*/20); + std::unique_ptr output_data_decode( + new int16_t[kPackets * decode_samples_per_channel * channels_]); + OpusRepacketizer* rp = opus_repacketizer_create(); + + size_t num_packets = 0; + constexpr size_t kMaxCycles = 100; + for (size_t idx = 0; idx < kMaxCycles; ++idx) { + auto speech_block = speech_data_.GetNextBlock(); + encoded_bytes_ = + WebRtcOpus_Encode(opus_encoder_, speech_block.data(), + rtc::CheckedDivExact(speech_block.size(), channels_), + kMaxBytes, bitstream_); + if (opus_repacketizer_cat(rp, bitstream_, + rtc::checked_cast(encoded_bytes_)) == + OPUS_OK) { + ++num_packets; + if (num_packets == kPackets) { + break; + } + } else { + // Opus repacketizer cannot guarantee a success. We try again if it fails. + opus_repacketizer_init(rp); + num_packets = 0; + } + } + EXPECT_EQ(kPackets, num_packets); + + encoded_bytes_ = opus_repacketizer_out(rp, bitstream_, kMaxBytes); + + EXPECT_EQ(decode_samples_per_channel * kPackets, + static_cast(WebRtcOpus_DurationEst( + opus_decoder_, bitstream_, encoded_bytes_))); + + EXPECT_EQ(decode_samples_per_channel * kPackets, + static_cast( + WebRtcOpus_Decode(opus_decoder_, bitstream_, encoded_bytes_, + output_data_decode.get(), &audio_type))); + + // Free memory. + opus_repacketizer_destroy(rp); + EXPECT_EQ(0, WebRtcOpus_EncoderFree(opus_encoder_)); + EXPECT_EQ(0, WebRtcOpus_DecoderFree(opus_decoder_)); +} + +TEST(OpusVadTest, CeltUnknownStatus) { + const uint8_t celt[] = {0x80}; + EXPECT_EQ(WebRtcOpus_PacketHasVoiceActivity(celt, 1), -1); +} + +TEST(OpusVadTest, Mono20msVadSet) { + uint8_t silk20msMonoVad[] = {0x78, 0x80}; + EXPECT_TRUE(WebRtcOpus_PacketHasVoiceActivity(silk20msMonoVad, 2)); +} + +TEST(OpusVadTest, Mono20MsVadUnset) { + uint8_t silk20msMonoSilence[] = {0x78, 0x00}; + EXPECT_FALSE(WebRtcOpus_PacketHasVoiceActivity(silk20msMonoSilence, 2)); +} + +TEST(OpusVadTest, Stereo20MsVadOnSideChannel) { + uint8_t silk20msStereoVadSideChannel[] = {0x78 | 0x04, 0x20}; + EXPECT_TRUE( + WebRtcOpus_PacketHasVoiceActivity(silk20msStereoVadSideChannel, 2)); +} + +TEST(OpusVadTest, TwoOpusMonoFramesVadOnSecond) { + uint8_t twoMonoFrames[] = {0x78 | 0x1, 0x00, 0x80}; + EXPECT_TRUE(WebRtcOpus_PacketHasVoiceActivity(twoMonoFrames, 3)); +} + +} // namespace webrtc diff --git a/third_party/libwebrtc/modules/audio_coding/codecs/opus/test/BUILD.gn b/third_party/libwebrtc/modules/audio_coding/codecs/opus/test/BUILD.gn new file mode 100644 index 0000000000..8bc0bf5e0e --- /dev/null +++ b/third_party/libwebrtc/modules/audio_coding/codecs/opus/test/BUILD.gn @@ -0,0 +1,55 @@ +# Copyright (c) 2018 The WebRTC project authors. All Rights Reserved. +# +# Use of this source code is governed by a BSD-style license +# that can be found in the LICENSE file in the root of the source +# tree. An additional intellectual property rights grant can be found +# in the file PATENTS. All contributing project authors may +# be found in the AUTHORS file in the root of the source tree. + +import("../../../../../webrtc.gni") + +visibility = [ + ":*", + "../../../:*", +] + +if (rtc_include_tests) { + rtc_library("test") { + testonly = true + + sources = [ + "audio_ring_buffer.cc", + "audio_ring_buffer.h", + "blocker.cc", + "blocker.h", + "lapped_transform.cc", + "lapped_transform.h", + ] + + deps = [ + "../../../../../common_audio", + "../../../../../common_audio:common_audio_c", + "../../../../../rtc_base:checks", + "../../../../../rtc_base/memory:aligned_malloc", + ] + } + + rtc_library("test_unittest") { + testonly = true + + sources = [ + "audio_ring_buffer_unittest.cc", + "blocker_unittest.cc", + "lapped_transform_unittest.cc", + ] + + deps = [ + ":test", + "../../../../../common_audio", + "../../../../../common_audio:common_audio_c", + "../../../../../rtc_base:macromagic", + "../../../../../test:test_support", + "//testing/gtest", + ] + } +} diff --git a/third_party/libwebrtc/modules/audio_coding/codecs/opus/test/audio_ring_buffer.cc b/third_party/libwebrtc/modules/audio_coding/codecs/opus/test/audio_ring_buffer.cc new file mode 100644 index 0000000000..2a71b43d2c --- /dev/null +++ b/third_party/libwebrtc/modules/audio_coding/codecs/opus/test/audio_ring_buffer.cc @@ -0,0 +1,76 @@ +/* + * Copyright (c) 2015 The WebRTC project authors. All Rights Reserved. + * + * Use of this source code is governed by a BSD-style license + * that can be found in the LICENSE file in the root of the source + * tree. An additional intellectual property rights grant can be found + * in the file PATENTS. All contributing project authors may + * be found in the AUTHORS file in the root of the source tree. + */ + +#include "modules/audio_coding/codecs/opus/test/audio_ring_buffer.h" + +#include "common_audio/ring_buffer.h" +#include "rtc_base/checks.h" + +// This is a simple multi-channel wrapper over the ring_buffer.h C interface. + +namespace webrtc { + +AudioRingBuffer::AudioRingBuffer(size_t channels, size_t max_frames) { + buffers_.reserve(channels); + for (size_t i = 0; i < channels; ++i) + buffers_.push_back(WebRtc_CreateBuffer(max_frames, sizeof(float))); +} + +AudioRingBuffer::~AudioRingBuffer() { + for (auto* buf : buffers_) + WebRtc_FreeBuffer(buf); +} + +void AudioRingBuffer::Write(const float* const* data, + size_t channels, + size_t frames) { + RTC_DCHECK_EQ(buffers_.size(), channels); + for (size_t i = 0; i < channels; ++i) { + const size_t written = WebRtc_WriteBuffer(buffers_[i], data[i], frames); + RTC_CHECK_EQ(written, frames); + } +} + +void AudioRingBuffer::Read(float* const* data, size_t channels, size_t frames) { + RTC_DCHECK_EQ(buffers_.size(), channels); + for (size_t i = 0; i < channels; ++i) { + const size_t read = + WebRtc_ReadBuffer(buffers_[i], nullptr, data[i], frames); + RTC_CHECK_EQ(read, frames); + } +} + +size_t AudioRingBuffer::ReadFramesAvailable() const { + // All buffers have the same amount available. + return WebRtc_available_read(buffers_[0]); +} + +size_t AudioRingBuffer::WriteFramesAvailable() const { + // All buffers have the same amount available. + return WebRtc_available_write(buffers_[0]); +} + +void AudioRingBuffer::MoveReadPositionForward(size_t frames) { + for (auto* buf : buffers_) { + const size_t moved = + static_cast(WebRtc_MoveReadPtr(buf, static_cast(frames))); + RTC_CHECK_EQ(moved, frames); + } +} + +void AudioRingBuffer::MoveReadPositionBackward(size_t frames) { + for (auto* buf : buffers_) { + const size_t moved = static_cast( + -WebRtc_MoveReadPtr(buf, -static_cast(frames))); + RTC_CHECK_EQ(moved, frames); + } +} + +} // namespace webrtc diff --git a/third_party/libwebrtc/modules/audio_coding/codecs/opus/test/audio_ring_buffer.h b/third_party/libwebrtc/modules/audio_coding/codecs/opus/test/audio_ring_buffer.h new file mode 100644 index 0000000000..a280ca2410 --- /dev/null +++ b/third_party/libwebrtc/modules/audio_coding/codecs/opus/test/audio_ring_buffer.h @@ -0,0 +1,57 @@ +/* + * Copyright (c) 2015 The WebRTC project authors. All Rights Reserved. + * + * Use of this source code is governed by a BSD-style license + * that can be found in the LICENSE file in the root of the source + * tree. An additional intellectual property rights grant can be found + * in the file PATENTS. All contributing project authors may + * be found in the AUTHORS file in the root of the source tree. + */ +#ifndef MODULES_AUDIO_CODING_CODECS_OPUS_TEST_AUDIO_RING_BUFFER_H_ +#define MODULES_AUDIO_CODING_CODECS_OPUS_TEST_AUDIO_RING_BUFFER_H_ + +#include + +#include +#include + +struct RingBuffer; + +namespace webrtc { + +// A ring buffer tailored for float deinterleaved audio. Any operation that +// cannot be performed as requested will cause a crash (e.g. insufficient data +// in the buffer to fulfill a read request.) +class AudioRingBuffer final { + public: + // Specify the number of channels and maximum number of frames the buffer will + // contain. + AudioRingBuffer(size_t channels, size_t max_frames); + ~AudioRingBuffer(); + + // Copies `data` to the buffer and advances the write pointer. `channels` must + // be the same as at creation time. + void Write(const float* const* data, size_t channels, size_t frames); + + // Copies from the buffer to `data` and advances the read pointer. `channels` + // must be the same as at creation time. + void Read(float* const* data, size_t channels, size_t frames); + + size_t ReadFramesAvailable() const; + size_t WriteFramesAvailable() const; + + // Moves the read position. The forward version advances the read pointer + // towards the write pointer and the backward verison withdraws the read + // pointer away from the write pointer (i.e. flushing and stuffing the buffer + // respectively.) + void MoveReadPositionForward(size_t frames); + void MoveReadPositionBackward(size_t frames); + + private: + // TODO(kwiberg): Use std::vector> instead. + std::vector buffers_; +}; + +} // namespace webrtc + +#endif // MODULES_AUDIO_CODING_CODECS_OPUS_TEST_AUDIO_RING_BUFFER_H_ diff --git a/third_party/libwebrtc/modules/audio_coding/codecs/opus/test/audio_ring_buffer_unittest.cc b/third_party/libwebrtc/modules/audio_coding/codecs/opus/test/audio_ring_buffer_unittest.cc new file mode 100644 index 0000000000..6dbc8ee9fe --- /dev/null +++ b/third_party/libwebrtc/modules/audio_coding/codecs/opus/test/audio_ring_buffer_unittest.cc @@ -0,0 +1,111 @@ +/* + * Copyright (c) 2015 The WebRTC project authors. All Rights Reserved. + * + * Use of this source code is governed by a BSD-style license + * that can be found in the LICENSE file in the root of the source + * tree. An additional intellectual property rights grant can be found + * in the file PATENTS. All contributing project authors may + * be found in the AUTHORS file in the root of the source tree. + */ + +#include "modules/audio_coding/codecs/opus/test/audio_ring_buffer.h" + +#include + +#include "common_audio/channel_buffer.h" +#include "test/gtest.h" + +namespace webrtc { + +class AudioRingBufferTest + : public ::testing::TestWithParam< ::testing::tuple > { +}; + +void ReadAndWriteTest(const ChannelBuffer& input, + size_t num_write_chunk_frames, + size_t num_read_chunk_frames, + size_t buffer_frames, + ChannelBuffer* output) { + const size_t num_channels = input.num_channels(); + const size_t total_frames = input.num_frames(); + AudioRingBuffer buf(num_channels, buffer_frames); + std::unique_ptr slice(new float*[num_channels]); + + size_t input_pos = 0; + size_t output_pos = 0; + while (input_pos + buf.WriteFramesAvailable() < total_frames) { + // Write until the buffer is as full as possible. + while (buf.WriteFramesAvailable() >= num_write_chunk_frames) { + buf.Write(input.Slice(slice.get(), input_pos), num_channels, + num_write_chunk_frames); + input_pos += num_write_chunk_frames; + } + // Read until the buffer is as empty as possible. + while (buf.ReadFramesAvailable() >= num_read_chunk_frames) { + EXPECT_LT(output_pos, total_frames); + buf.Read(output->Slice(slice.get(), output_pos), num_channels, + num_read_chunk_frames); + output_pos += num_read_chunk_frames; + } + } + + // Write and read the last bit. + if (input_pos < total_frames) { + buf.Write(input.Slice(slice.get(), input_pos), num_channels, + total_frames - input_pos); + } + if (buf.ReadFramesAvailable()) { + buf.Read(output->Slice(slice.get(), output_pos), num_channels, + buf.ReadFramesAvailable()); + } + EXPECT_EQ(0u, buf.ReadFramesAvailable()); +} + +TEST_P(AudioRingBufferTest, ReadDataMatchesWrittenData) { + const size_t kFrames = 5000; + const size_t num_channels = ::testing::get<3>(GetParam()); + + // Initialize the input data to an increasing sequence. + ChannelBuffer input(kFrames, static_cast(num_channels)); + for (size_t i = 0; i < num_channels; ++i) + for (size_t j = 0; j < kFrames; ++j) + input.channels()[i][j] = (i + 1) * (j + 1); + + ChannelBuffer output(kFrames, static_cast(num_channels)); + ReadAndWriteTest(input, ::testing::get<0>(GetParam()), + ::testing::get<1>(GetParam()), ::testing::get<2>(GetParam()), + &output); + + // Verify the read data matches the input. + for (size_t i = 0; i < num_channels; ++i) + for (size_t j = 0; j < kFrames; ++j) + EXPECT_EQ(input.channels()[i][j], output.channels()[i][j]); +} + +INSTANTIATE_TEST_SUITE_P( + AudioRingBufferTest, + AudioRingBufferTest, + ::testing::Combine(::testing::Values(10, 20, 42), // num_write_chunk_frames + ::testing::Values(1, 10, 17), // num_read_chunk_frames + ::testing::Values(100, 256), // buffer_frames + ::testing::Values(1, 4))); // num_channels + +TEST_F(AudioRingBufferTest, MoveReadPosition) { + const size_t kNumChannels = 1; + const float kInputArray[] = {1, 2, 3, 4}; + const size_t kNumFrames = sizeof(kInputArray) / sizeof(*kInputArray); + ChannelBuffer input(kNumFrames, kNumChannels); + input.SetDataForTesting(kInputArray, kNumFrames); + AudioRingBuffer buf(kNumChannels, kNumFrames); + buf.Write(input.channels(), kNumChannels, kNumFrames); + + buf.MoveReadPositionForward(3); + ChannelBuffer output(1, kNumChannels); + buf.Read(output.channels(), kNumChannels, 1); + EXPECT_EQ(4, output.channels()[0][0]); + buf.MoveReadPositionBackward(3); + buf.Read(output.channels(), kNumChannels, 1); + EXPECT_EQ(2, output.channels()[0][0]); +} + +} // namespace webrtc diff --git a/third_party/libwebrtc/modules/audio_coding/codecs/opus/test/blocker.cc b/third_party/libwebrtc/modules/audio_coding/codecs/opus/test/blocker.cc new file mode 100644 index 0000000000..33406cead9 --- /dev/null +++ b/third_party/libwebrtc/modules/audio_coding/codecs/opus/test/blocker.cc @@ -0,0 +1,215 @@ +/* + * Copyright (c) 2014 The WebRTC project authors. All Rights Reserved. + * + * Use of this source code is governed by a BSD-style license + * that can be found in the LICENSE file in the root of the source + * tree. An additional intellectual property rights grant can be found + * in the file PATENTS. All contributing project authors may + * be found in the AUTHORS file in the root of the source tree. + */ + +#include "modules/audio_coding/codecs/opus/test/blocker.h" + +#include + +#include "rtc_base/checks.h" + +namespace { + +// Adds `a` and `b` frame by frame into `result` (basically matrix addition). +void AddFrames(const float* const* a, + size_t a_start_index, + const float* const* b, + int b_start_index, + size_t num_frames, + size_t num_channels, + float* const* result, + size_t result_start_index) { + for (size_t i = 0; i < num_channels; ++i) { + for (size_t j = 0; j < num_frames; ++j) { + result[i][j + result_start_index] = + a[i][j + a_start_index] + b[i][j + b_start_index]; + } + } +} + +// Copies `src` into `dst` channel by channel. +void CopyFrames(const float* const* src, + size_t src_start_index, + size_t num_frames, + size_t num_channels, + float* const* dst, + size_t dst_start_index) { + for (size_t i = 0; i < num_channels; ++i) { + memcpy(&dst[i][dst_start_index], &src[i][src_start_index], + num_frames * sizeof(dst[i][dst_start_index])); + } +} + +// Moves `src` into `dst` channel by channel. +void MoveFrames(const float* const* src, + size_t src_start_index, + size_t num_frames, + size_t num_channels, + float* const* dst, + size_t dst_start_index) { + for (size_t i = 0; i < num_channels; ++i) { + memmove(&dst[i][dst_start_index], &src[i][src_start_index], + num_frames * sizeof(dst[i][dst_start_index])); + } +} + +void ZeroOut(float* const* buffer, + size_t starting_idx, + size_t num_frames, + size_t num_channels) { + for (size_t i = 0; i < num_channels; ++i) { + memset(&buffer[i][starting_idx], 0, + num_frames * sizeof(buffer[i][starting_idx])); + } +} + +// Pointwise multiplies each channel of `frames` with `window`. Results are +// stored in `frames`. +void ApplyWindow(const float* window, + size_t num_frames, + size_t num_channels, + float* const* frames) { + for (size_t i = 0; i < num_channels; ++i) { + for (size_t j = 0; j < num_frames; ++j) { + frames[i][j] = frames[i][j] * window[j]; + } + } +} + +size_t gcd(size_t a, size_t b) { + size_t tmp; + while (b) { + tmp = a; + a = b; + b = tmp % b; + } + return a; +} + +} // namespace + +namespace webrtc { + +Blocker::Blocker(size_t chunk_size, + size_t block_size, + size_t num_input_channels, + size_t num_output_channels, + const float* window, + size_t shift_amount, + BlockerCallback* callback) + : chunk_size_(chunk_size), + block_size_(block_size), + num_input_channels_(num_input_channels), + num_output_channels_(num_output_channels), + initial_delay_(block_size_ - gcd(chunk_size, shift_amount)), + frame_offset_(0), + input_buffer_(num_input_channels_, chunk_size_ + initial_delay_), + output_buffer_(chunk_size_ + initial_delay_, num_output_channels_), + input_block_(block_size_, num_input_channels_), + output_block_(block_size_, num_output_channels_), + window_(new float[block_size_]), + shift_amount_(shift_amount), + callback_(callback) { + RTC_CHECK_LE(num_output_channels_, num_input_channels_); + RTC_CHECK_LE(shift_amount_, block_size_); + + memcpy(window_.get(), window, block_size_ * sizeof(*window_.get())); + input_buffer_.MoveReadPositionBackward(initial_delay_); +} + +Blocker::~Blocker() = default; + +// When block_size < chunk_size the input and output buffers look like this: +// +// delay* chunk_size chunk_size + delay* +// buffer: <-------------|---------------------|---------------|> +// _a_ _b_ _c_ +// +// On each call to ProcessChunk(): +// 1. New input gets read into sections _b_ and _c_ of the input buffer. +// 2. We block starting from frame_offset. +// 3. We block until we reach a block `bl` that doesn't contain any frames +// from sections _a_ or _b_ of the input buffer. +// 4. We window the current block, fire the callback for processing, window +// again, and overlap/add to the output buffer. +// 5. We copy sections _a_ and _b_ of the output buffer into output. +// 6. For both the input and the output buffers, we copy section _c_ into +// section _a_. +// 7. We set the new frame_offset to be the difference between the first frame +// of `bl` and the border between sections _b_ and _c_. +// +// When block_size > chunk_size the input and output buffers look like this: +// +// chunk_size delay* chunk_size + delay* +// buffer: <-------------|---------------------|---------------|> +// _a_ _b_ _c_ +// +// On each call to ProcessChunk(): +// The procedure is the same as above, except for: +// 1. New input gets read into section _c_ of the input buffer. +// 3. We block until we reach a block `bl` that doesn't contain any frames +// from section _a_ of the input buffer. +// 5. We copy section _a_ of the output buffer into output. +// 6. For both the input and the output buffers, we copy sections _b_ and _c_ +// into section _a_ and _b_. +// 7. We set the new frame_offset to be the difference between the first frame +// of `bl` and the border between sections _a_ and _b_. +// +// * delay here refers to inintial_delay_ +// +// TODO(claguna): Look at using ring buffers to eliminate some copies. +void Blocker::ProcessChunk(const float* const* input, + size_t chunk_size, + size_t num_input_channels, + size_t num_output_channels, + float* const* output) { + RTC_CHECK_EQ(chunk_size, chunk_size_); + RTC_CHECK_EQ(num_input_channels, num_input_channels_); + RTC_CHECK_EQ(num_output_channels, num_output_channels_); + + input_buffer_.Write(input, num_input_channels, chunk_size_); + size_t first_frame_in_block = frame_offset_; + + // Loop through blocks. + while (first_frame_in_block < chunk_size_) { + input_buffer_.Read(input_block_.channels(), num_input_channels, + block_size_); + input_buffer_.MoveReadPositionBackward(block_size_ - shift_amount_); + + ApplyWindow(window_.get(), block_size_, num_input_channels_, + input_block_.channels()); + callback_->ProcessBlock(input_block_.channels(), block_size_, + num_input_channels_, num_output_channels_, + output_block_.channels()); + ApplyWindow(window_.get(), block_size_, num_output_channels_, + output_block_.channels()); + + AddFrames(output_buffer_.channels(), first_frame_in_block, + output_block_.channels(), 0, block_size_, num_output_channels_, + output_buffer_.channels(), first_frame_in_block); + + first_frame_in_block += shift_amount_; + } + + // Copy output buffer to output + CopyFrames(output_buffer_.channels(), 0, chunk_size_, num_output_channels_, + output, 0); + + // Copy output buffer [chunk_size_, chunk_size_ + initial_delay] + // to output buffer [0, initial_delay], zero the rest. + MoveFrames(output_buffer_.channels(), chunk_size, initial_delay_, + num_output_channels_, output_buffer_.channels(), 0); + ZeroOut(output_buffer_.channels(), initial_delay_, chunk_size_, + num_output_channels_); + + // Calculate new starting frames. + frame_offset_ = first_frame_in_block - chunk_size_; +} + +} // namespace webrtc diff --git a/third_party/libwebrtc/modules/audio_coding/codecs/opus/test/blocker.h b/third_party/libwebrtc/modules/audio_coding/codecs/opus/test/blocker.h new file mode 100644 index 0000000000..59b7e29621 --- /dev/null +++ b/third_party/libwebrtc/modules/audio_coding/codecs/opus/test/blocker.h @@ -0,0 +1,127 @@ +/* + * Copyright (c) 2014 The WebRTC project authors. All Rights Reserved. + * + * Use of this source code is governed by a BSD-style license + * that can be found in the LICENSE file in the root of the source + * tree. An additional intellectual property rights grant can be found + * in the file PATENTS. All contributing project authors may + * be found in the AUTHORS file in the root of the source tree. + */ + +#ifndef MODULES_AUDIO_CODING_CODECS_OPUS_TEST_BLOCKER_H_ +#define MODULES_AUDIO_CODING_CODECS_OPUS_TEST_BLOCKER_H_ + +#include + +#include "common_audio/channel_buffer.h" +#include "modules/audio_coding/codecs/opus/test/audio_ring_buffer.h" + +namespace webrtc { + +// The callback function to process audio in the time domain. Input has already +// been windowed, and output will be windowed. The number of input channels +// must be >= the number of output channels. +class BlockerCallback { + public: + virtual ~BlockerCallback() {} + + virtual void ProcessBlock(const float* const* input, + size_t num_frames, + size_t num_input_channels, + size_t num_output_channels, + float* const* output) = 0; +}; + +// The main purpose of Blocker is to abstract away the fact that often we +// receive a different number of audio frames than our transform takes. For +// example, most FFTs work best when the fft-size is a power of 2, but suppose +// we receive 20ms of audio at a sample rate of 48000. That comes to 960 frames +// of audio, which is not a power of 2. Blocker allows us to specify the +// transform and all other necessary processing via the Process() callback +// function without any constraints on the transform-size +// (read: `block_size_`) or received-audio-size (read: `chunk_size_`). +// We handle this for the multichannel audio case, allowing for different +// numbers of input and output channels (for example, beamforming takes 2 or +// more input channels and returns 1 output channel). Audio signals are +// represented as deinterleaved floats in the range [-1, 1]. +// +// Blocker is responsible for: +// - blocking audio while handling potential discontinuities on the edges +// of chunks +// - windowing blocks before sending them to Process() +// - windowing processed blocks, and overlap-adding them together before +// sending back a processed chunk +// +// To use blocker: +// 1. Impelment a BlockerCallback object `bc`. +// 2. Instantiate a Blocker object `b`, passing in `bc`. +// 3. As you receive audio, call b.ProcessChunk() to get processed audio. +// +// A small amount of delay is added to the first received chunk to deal with +// the difference in chunk/block sizes. This delay is <= chunk_size. +// +// Ownership of window is retained by the caller. That is, Blocker makes a +// copy of window and does not attempt to delete it. +class Blocker { + public: + Blocker(size_t chunk_size, + size_t block_size, + size_t num_input_channels, + size_t num_output_channels, + const float* window, + size_t shift_amount, + BlockerCallback* callback); + ~Blocker(); + + void ProcessChunk(const float* const* input, + size_t chunk_size, + size_t num_input_channels, + size_t num_output_channels, + float* const* output); + + size_t initial_delay() const { return initial_delay_; } + + private: + const size_t chunk_size_; + const size_t block_size_; + const size_t num_input_channels_; + const size_t num_output_channels_; + + // The number of frames of delay to add at the beginning of the first chunk. + const size_t initial_delay_; + + // The frame index into the input buffer where the first block should be read + // from. This is necessary because shift_amount_ is not necessarily a + // multiple of chunk_size_, so blocks won't line up at the start of the + // buffer. + size_t frame_offset_; + + // Since blocks nearly always overlap, there are certain blocks that require + // frames from the end of one chunk and the beginning of the next chunk. The + // input and output buffers are responsible for saving those frames between + // calls to ProcessChunk(). + // + // Both contain |initial delay| + `chunk_size` frames. The input is a fairly + // standard FIFO, but due to the overlap-add it's harder to use an + // AudioRingBuffer for the output. + AudioRingBuffer input_buffer_; + ChannelBuffer output_buffer_; + + // Space for the input block (can't wrap because of windowing). + ChannelBuffer input_block_; + + // Space for the output block (can't wrap because of overlap/add). + ChannelBuffer output_block_; + + std::unique_ptr window_; + + // The amount of frames between the start of contiguous blocks. For example, + // `shift_amount_` = `block_size_` / 2 for a Hann window. + size_t shift_amount_; + + BlockerCallback* callback_; +}; + +} // namespace webrtc + +#endif // MODULES_AUDIO_CODING_CODECS_OPUS_TEST_BLOCKER_H_ diff --git a/third_party/libwebrtc/modules/audio_coding/codecs/opus/test/blocker_unittest.cc b/third_party/libwebrtc/modules/audio_coding/codecs/opus/test/blocker_unittest.cc new file mode 100644 index 0000000000..9c8e789ba9 --- /dev/null +++ b/third_party/libwebrtc/modules/audio_coding/codecs/opus/test/blocker_unittest.cc @@ -0,0 +1,293 @@ +/* + * Copyright (c) 2014 The WebRTC project authors. All Rights Reserved. + * + * Use of this source code is governed by a BSD-style license + * that can be found in the LICENSE file in the root of the source + * tree. An additional intellectual property rights grant can be found + * in the file PATENTS. All contributing project authors may + * be found in the AUTHORS file in the root of the source tree. + */ + +#include "modules/audio_coding/codecs/opus/test/blocker.h" + +#include + +#include "rtc_base/arraysize.h" +#include "test/gtest.h" + +namespace { + +// Callback Function to add 3 to every sample in the signal. +class PlusThreeBlockerCallback : public webrtc::BlockerCallback { + public: + void ProcessBlock(const float* const* input, + size_t num_frames, + size_t num_input_channels, + size_t num_output_channels, + float* const* output) override { + for (size_t i = 0; i < num_output_channels; ++i) { + for (size_t j = 0; j < num_frames; ++j) { + output[i][j] = input[i][j] + 3; + } + } + } +}; + +// No-op Callback Function. +class CopyBlockerCallback : public webrtc::BlockerCallback { + public: + void ProcessBlock(const float* const* input, + size_t num_frames, + size_t num_input_channels, + size_t num_output_channels, + float* const* output) override { + for (size_t i = 0; i < num_output_channels; ++i) { + for (size_t j = 0; j < num_frames; ++j) { + output[i][j] = input[i][j]; + } + } + } +}; + +} // namespace + +namespace webrtc { + +// Tests blocking with a window that multiplies the signal by 2, a callback +// that adds 3 to each sample in the signal, and different combinations of chunk +// size, block size, and shift amount. +class BlockerTest : public ::testing::Test { + protected: + void RunTest(Blocker* blocker, + size_t chunk_size, + size_t num_frames, + const float* const* input, + float* const* input_chunk, + float* const* output, + float* const* output_chunk, + size_t num_input_channels, + size_t num_output_channels) { + size_t start = 0; + size_t end = chunk_size - 1; + while (end < num_frames) { + CopyTo(input_chunk, 0, start, num_input_channels, chunk_size, input); + blocker->ProcessChunk(input_chunk, chunk_size, num_input_channels, + num_output_channels, output_chunk); + CopyTo(output, start, 0, num_output_channels, chunk_size, output_chunk); + + start += chunk_size; + end += chunk_size; + } + } + + void ValidateSignalEquality(const float* const* expected, + const float* const* actual, + size_t num_channels, + size_t num_frames) { + for (size_t i = 0; i < num_channels; ++i) { + for (size_t j = 0; j < num_frames; ++j) { + EXPECT_FLOAT_EQ(expected[i][j], actual[i][j]); + } + } + } + + void ValidateInitialDelay(const float* const* output, + size_t num_channels, + size_t num_frames, + size_t initial_delay) { + for (size_t i = 0; i < num_channels; ++i) { + for (size_t j = 0; j < num_frames; ++j) { + if (j < initial_delay) { + EXPECT_FLOAT_EQ(output[i][j], 0.f); + } else { + EXPECT_GT(output[i][j], 0.f); + } + } + } + } + + static void CopyTo(float* const* dst, + size_t start_index_dst, + size_t start_index_src, + size_t num_channels, + size_t num_frames, + const float* const* src) { + for (size_t i = 0; i < num_channels; ++i) { + memcpy(&dst[i][start_index_dst], &src[i][start_index_src], + num_frames * sizeof(float)); + } + } +}; + +TEST_F(BlockerTest, TestBlockerMutuallyPrimeChunkandBlockSize) { + const size_t kNumInputChannels = 3; + const size_t kNumOutputChannels = 2; + const size_t kNumFrames = 10; + const size_t kBlockSize = 4; + const size_t kChunkSize = 5; + const size_t kShiftAmount = 2; + + const float kInput[kNumInputChannels][kNumFrames] = { + {1, 1, 1, 1, 1, 1, 1, 1, 1, 1}, + {2, 2, 2, 2, 2, 2, 2, 2, 2, 2}, + {3, 3, 3, 3, 3, 3, 3, 3, 3, 3}}; + ChannelBuffer input_cb(kNumFrames, kNumInputChannels); + input_cb.SetDataForTesting(kInput[0], sizeof(kInput) / sizeof(**kInput)); + + const float kExpectedOutput[kNumInputChannels][kNumFrames] = { + {6, 6, 12, 20, 20, 20, 20, 20, 20, 20}, + {6, 6, 12, 28, 28, 28, 28, 28, 28, 28}}; + ChannelBuffer expected_output_cb(kNumFrames, kNumInputChannels); + expected_output_cb.SetDataForTesting( + kExpectedOutput[0], sizeof(kExpectedOutput) / sizeof(**kExpectedOutput)); + + const float kWindow[kBlockSize] = {2.f, 2.f, 2.f, 2.f}; + + ChannelBuffer actual_output_cb(kNumFrames, kNumOutputChannels); + ChannelBuffer input_chunk_cb(kChunkSize, kNumInputChannels); + ChannelBuffer output_chunk_cb(kChunkSize, kNumOutputChannels); + + PlusThreeBlockerCallback callback; + Blocker blocker(kChunkSize, kBlockSize, kNumInputChannels, kNumOutputChannels, + kWindow, kShiftAmount, &callback); + + RunTest(&blocker, kChunkSize, kNumFrames, input_cb.channels(), + input_chunk_cb.channels(), actual_output_cb.channels(), + output_chunk_cb.channels(), kNumInputChannels, kNumOutputChannels); + + ValidateSignalEquality(expected_output_cb.channels(), + actual_output_cb.channels(), kNumOutputChannels, + kNumFrames); +} + +TEST_F(BlockerTest, TestBlockerMutuallyPrimeShiftAndBlockSize) { + const size_t kNumInputChannels = 3; + const size_t kNumOutputChannels = 2; + const size_t kNumFrames = 12; + const size_t kBlockSize = 4; + const size_t kChunkSize = 6; + const size_t kShiftAmount = 3; + + const float kInput[kNumInputChannels][kNumFrames] = { + {1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1}, + {2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2}, + {3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3}}; + ChannelBuffer input_cb(kNumFrames, kNumInputChannels); + input_cb.SetDataForTesting(kInput[0], sizeof(kInput) / sizeof(**kInput)); + + const float kExpectedOutput[kNumOutputChannels][kNumFrames] = { + {6, 10, 10, 20, 10, 10, 20, 10, 10, 20, 10, 10}, + {6, 14, 14, 28, 14, 14, 28, 14, 14, 28, 14, 14}}; + ChannelBuffer expected_output_cb(kNumFrames, kNumOutputChannels); + expected_output_cb.SetDataForTesting( + kExpectedOutput[0], sizeof(kExpectedOutput) / sizeof(**kExpectedOutput)); + + const float kWindow[kBlockSize] = {2.f, 2.f, 2.f, 2.f}; + + ChannelBuffer actual_output_cb(kNumFrames, kNumOutputChannels); + ChannelBuffer input_chunk_cb(kChunkSize, kNumInputChannels); + ChannelBuffer output_chunk_cb(kChunkSize, kNumOutputChannels); + + PlusThreeBlockerCallback callback; + Blocker blocker(kChunkSize, kBlockSize, kNumInputChannels, kNumOutputChannels, + kWindow, kShiftAmount, &callback); + + RunTest(&blocker, kChunkSize, kNumFrames, input_cb.channels(), + input_chunk_cb.channels(), actual_output_cb.channels(), + output_chunk_cb.channels(), kNumInputChannels, kNumOutputChannels); + + ValidateSignalEquality(expected_output_cb.channels(), + actual_output_cb.channels(), kNumOutputChannels, + kNumFrames); +} + +TEST_F(BlockerTest, TestBlockerNoOverlap) { + const size_t kNumInputChannels = 3; + const size_t kNumOutputChannels = 2; + const size_t kNumFrames = 12; + const size_t kBlockSize = 4; + const size_t kChunkSize = 4; + const size_t kShiftAmount = 4; + + const float kInput[kNumInputChannels][kNumFrames] = { + {1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1}, + {2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2}, + {3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3}}; + ChannelBuffer input_cb(kNumFrames, kNumInputChannels); + input_cb.SetDataForTesting(kInput[0], sizeof(kInput) / sizeof(**kInput)); + + const float kExpectedOutput[kNumOutputChannels][kNumFrames] = { + {10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10}, + {14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14}}; + ChannelBuffer expected_output_cb(kNumFrames, kNumOutputChannels); + expected_output_cb.SetDataForTesting( + kExpectedOutput[0], sizeof(kExpectedOutput) / sizeof(**kExpectedOutput)); + + const float kWindow[kBlockSize] = {2.f, 2.f, 2.f, 2.f}; + + ChannelBuffer actual_output_cb(kNumFrames, kNumOutputChannels); + ChannelBuffer input_chunk_cb(kChunkSize, kNumInputChannels); + ChannelBuffer output_chunk_cb(kChunkSize, kNumOutputChannels); + + PlusThreeBlockerCallback callback; + Blocker blocker(kChunkSize, kBlockSize, kNumInputChannels, kNumOutputChannels, + kWindow, kShiftAmount, &callback); + + RunTest(&blocker, kChunkSize, kNumFrames, input_cb.channels(), + input_chunk_cb.channels(), actual_output_cb.channels(), + output_chunk_cb.channels(), kNumInputChannels, kNumOutputChannels); + + ValidateSignalEquality(expected_output_cb.channels(), + actual_output_cb.channels(), kNumOutputChannels, + kNumFrames); +} + +TEST_F(BlockerTest, InitialDelaysAreMinimum) { + const size_t kNumInputChannels = 3; + const size_t kNumOutputChannels = 2; + const size_t kNumFrames = 1280; + const size_t kChunkSize[] = {80, 80, 80, 80, 80, 80, + 160, 160, 160, 160, 160, 160}; + const size_t kBlockSize[] = {64, 64, 64, 128, 128, 128, + 128, 128, 128, 256, 256, 256}; + const size_t kShiftAmount[] = {16, 32, 64, 32, 64, 128, + 32, 64, 128, 64, 128, 256}; + const size_t kInitialDelay[] = {48, 48, 48, 112, 112, 112, + 96, 96, 96, 224, 224, 224}; + + float input[kNumInputChannels][kNumFrames]; + for (size_t i = 0; i < kNumInputChannels; ++i) { + for (size_t j = 0; j < kNumFrames; ++j) { + input[i][j] = i + 1; + } + } + ChannelBuffer input_cb(kNumFrames, kNumInputChannels); + input_cb.SetDataForTesting(input[0], sizeof(input) / sizeof(**input)); + + ChannelBuffer output_cb(kNumFrames, kNumOutputChannels); + + CopyBlockerCallback callback; + + for (size_t i = 0; i < arraysize(kChunkSize); ++i) { + std::unique_ptr window(new float[kBlockSize[i]]); + for (size_t j = 0; j < kBlockSize[i]; ++j) { + window[j] = 1.f; + } + + ChannelBuffer input_chunk_cb(kChunkSize[i], kNumInputChannels); + ChannelBuffer output_chunk_cb(kChunkSize[i], kNumOutputChannels); + + Blocker blocker(kChunkSize[i], kBlockSize[i], kNumInputChannels, + kNumOutputChannels, window.get(), kShiftAmount[i], + &callback); + + RunTest(&blocker, kChunkSize[i], kNumFrames, input_cb.channels(), + input_chunk_cb.channels(), output_cb.channels(), + output_chunk_cb.channels(), kNumInputChannels, kNumOutputChannels); + + ValidateInitialDelay(output_cb.channels(), kNumOutputChannels, kNumFrames, + kInitialDelay[i]); + } +} + +} // namespace webrtc diff --git a/third_party/libwebrtc/modules/audio_coding/codecs/opus/test/lapped_transform.cc b/third_party/libwebrtc/modules/audio_coding/codecs/opus/test/lapped_transform.cc new file mode 100644 index 0000000000..b1a6526bba --- /dev/null +++ b/third_party/libwebrtc/modules/audio_coding/codecs/opus/test/lapped_transform.cc @@ -0,0 +1,100 @@ +/* + * Copyright (c) 2014 The WebRTC project authors. All Rights Reserved. + * + * Use of this source code is governed by a BSD-style license + * that can be found in the LICENSE file in the root of the source + * tree. An additional intellectual property rights grant can be found + * in the file PATENTS. All contributing project authors may + * be found in the AUTHORS file in the root of the source tree. + */ + +#include "modules/audio_coding/codecs/opus/test/lapped_transform.h" + +#include +#include +#include + +#include "common_audio/real_fourier.h" +#include "rtc_base/checks.h" + +namespace webrtc { + +void LappedTransform::BlockThunk::ProcessBlock(const float* const* input, + size_t num_frames, + size_t num_input_channels, + size_t num_output_channels, + float* const* output) { + RTC_CHECK_EQ(num_input_channels, parent_->num_in_channels_); + RTC_CHECK_EQ(num_output_channels, parent_->num_out_channels_); + RTC_CHECK_EQ(parent_->block_length_, num_frames); + + for (size_t i = 0; i < num_input_channels; ++i) { + memcpy(parent_->real_buf_.Row(i), input[i], num_frames * sizeof(*input[0])); + parent_->fft_->Forward(parent_->real_buf_.Row(i), + parent_->cplx_pre_.Row(i)); + } + + size_t block_length = + RealFourier::ComplexLength(RealFourier::FftOrder(num_frames)); + RTC_CHECK_EQ(parent_->cplx_length_, block_length); + parent_->block_processor_->ProcessAudioBlock( + parent_->cplx_pre_.Array(), num_input_channels, parent_->cplx_length_, + num_output_channels, parent_->cplx_post_.Array()); + + for (size_t i = 0; i < num_output_channels; ++i) { + parent_->fft_->Inverse(parent_->cplx_post_.Row(i), + parent_->real_buf_.Row(i)); + memcpy(output[i], parent_->real_buf_.Row(i), + num_frames * sizeof(*input[0])); + } +} + +LappedTransform::LappedTransform(size_t num_in_channels, + size_t num_out_channels, + size_t chunk_length, + const float* window, + size_t block_length, + size_t shift_amount, + Callback* callback) + : blocker_callback_(this), + num_in_channels_(num_in_channels), + num_out_channels_(num_out_channels), + block_length_(block_length), + chunk_length_(chunk_length), + block_processor_(callback), + blocker_(chunk_length_, + block_length_, + num_in_channels_, + num_out_channels_, + window, + shift_amount, + &blocker_callback_), + fft_(RealFourier::Create(RealFourier::FftOrder(block_length_))), + cplx_length_(RealFourier::ComplexLength(fft_->order())), + real_buf_(num_in_channels, + block_length_, + RealFourier::kFftBufferAlignment), + cplx_pre_(num_in_channels, + cplx_length_, + RealFourier::kFftBufferAlignment), + cplx_post_(num_out_channels, + cplx_length_, + RealFourier::kFftBufferAlignment) { + RTC_CHECK(num_in_channels_ > 0); + RTC_CHECK_GT(block_length_, 0); + RTC_CHECK_GT(chunk_length_, 0); + RTC_CHECK(block_processor_); + + // block_length_ power of 2? + RTC_CHECK_EQ(0, block_length_ & (block_length_ - 1)); +} + +LappedTransform::~LappedTransform() = default; + +void LappedTransform::ProcessChunk(const float* const* in_chunk, + float* const* out_chunk) { + blocker_.ProcessChunk(in_chunk, chunk_length_, num_in_channels_, + num_out_channels_, out_chunk); +} + +} // namespace webrtc diff --git a/third_party/libwebrtc/modules/audio_coding/codecs/opus/test/lapped_transform.h b/third_party/libwebrtc/modules/audio_coding/codecs/opus/test/lapped_transform.h new file mode 100644 index 0000000000..bb25c34a9e --- /dev/null +++ b/third_party/libwebrtc/modules/audio_coding/codecs/opus/test/lapped_transform.h @@ -0,0 +1,175 @@ +/* + * Copyright (c) 2014 The WebRTC project authors. All Rights Reserved. + * + * Use of this source code is governed by a BSD-style license + * that can be found in the LICENSE file in the root of the source + * tree. An additional intellectual property rights grant can be found + * in the file PATENTS. All contributing project authors may + * be found in the AUTHORS file in the root of the source tree. + */ + +#ifndef MODULES_AUDIO_CODING_CODECS_OPUS_TEST_LAPPED_TRANSFORM_H_ +#define MODULES_AUDIO_CODING_CODECS_OPUS_TEST_LAPPED_TRANSFORM_H_ + +#include +#include + +#include "common_audio/real_fourier.h" +#include "modules/audio_coding/codecs/opus/test/blocker.h" +#include "rtc_base/memory/aligned_malloc.h" + +namespace webrtc { + +// Wrapper class for aligned arrays. Every row (and the first dimension) are +// aligned to the given byte alignment. +template +class AlignedArray { + public: + AlignedArray(size_t rows, size_t cols, size_t alignment) + : rows_(rows), cols_(cols) { + RTC_CHECK_GT(alignment, 0); + head_row_ = + static_cast(AlignedMalloc(rows_ * sizeof(*head_row_), alignment)); + for (size_t i = 0; i < rows_; ++i) { + head_row_[i] = static_cast( + AlignedMalloc(cols_ * sizeof(**head_row_), alignment)); + } + } + + ~AlignedArray() { + for (size_t i = 0; i < rows_; ++i) { + AlignedFree(head_row_[i]); + } + AlignedFree(head_row_); + } + + T* const* Array() { return head_row_; } + + const T* const* Array() const { return head_row_; } + + T* Row(size_t row) { + RTC_CHECK_LE(row, rows_); + return head_row_[row]; + } + + const T* Row(size_t row) const { + RTC_CHECK_LE(row, rows_); + return head_row_[row]; + } + + private: + size_t rows_; + size_t cols_; + T** head_row_; +}; + +// Helper class for audio processing modules which operate on frequency domain +// input derived from the windowed time domain audio stream. +// +// The input audio chunk is sliced into possibly overlapping blocks, multiplied +// by a window and transformed with an FFT implementation. The transformed data +// is supplied to the given callback for processing. The processed output is +// then inverse transformed into the time domain and spliced back into a chunk +// which constitutes the final output of this processing module. +class LappedTransform { + public: + class Callback { + public: + virtual ~Callback() {} + + virtual void ProcessAudioBlock(const std::complex* const* in_block, + size_t num_in_channels, + size_t frames, + size_t num_out_channels, + std::complex* const* out_block) = 0; + }; + + // Construct a transform instance. `chunk_length` is the number of samples in + // each channel. `window` defines the window, owned by the caller (a copy is + // made internally); `window` should have length equal to `block_length`. + // `block_length` defines the length of a block, in samples. + // `shift_amount` is in samples. `callback` is the caller-owned audio + // processing function called for each block of the input chunk. + LappedTransform(size_t num_in_channels, + size_t num_out_channels, + size_t chunk_length, + const float* window, + size_t block_length, + size_t shift_amount, + Callback* callback); + ~LappedTransform(); + + // Main audio processing helper method. Internally slices `in_chunk` into + // blocks, transforms them to frequency domain, calls the callback for each + // block and returns a de-blocked time domain chunk of audio through + // `out_chunk`. Both buffers are caller-owned. + void ProcessChunk(const float* const* in_chunk, float* const* out_chunk); + + // Get the chunk length. + // + // The chunk length is the number of samples per channel that must be passed + // to ProcessChunk via the parameter in_chunk. + // + // Returns the same chunk_length passed to the LappedTransform constructor. + size_t chunk_length() const { return chunk_length_; } + + // Get the number of input channels. + // + // This is the number of arrays that must be passed to ProcessChunk via + // in_chunk. + // + // Returns the same num_in_channels passed to the LappedTransform constructor. + size_t num_in_channels() const { return num_in_channels_; } + + // Get the number of output channels. + // + // This is the number of arrays that must be passed to ProcessChunk via + // out_chunk. + // + // Returns the same num_out_channels passed to the LappedTransform + // constructor. + size_t num_out_channels() const { return num_out_channels_; } + + // Returns the initial delay. + // + // This is the delay introduced by the `blocker_` to be able to get and return + // chunks of `chunk_length`, but process blocks of `block_length`. + size_t initial_delay() const { return blocker_.initial_delay(); } + + private: + // Internal middleware callback, given to the blocker. Transforms each block + // and hands it over to the processing method given at construction time. + class BlockThunk : public BlockerCallback { + public: + explicit BlockThunk(LappedTransform* parent) : parent_(parent) {} + + void ProcessBlock(const float* const* input, + size_t num_frames, + size_t num_input_channels, + size_t num_output_channels, + float* const* output) override; + + private: + LappedTransform* const parent_; + } blocker_callback_; + + const size_t num_in_channels_; + const size_t num_out_channels_; + + const size_t block_length_; + const size_t chunk_length_; + + Callback* const block_processor_; + Blocker blocker_; + + // TODO(alessiob): Replace RealFourier with a different FFT library. + std::unique_ptr fft_; + const size_t cplx_length_; + AlignedArray real_buf_; + AlignedArray > cplx_pre_; + AlignedArray > cplx_post_; +}; + +} // namespace webrtc + +#endif // MODULES_AUDIO_CODING_CODECS_OPUS_TEST_LAPPED_TRANSFORM_H_ diff --git a/third_party/libwebrtc/modules/audio_coding/codecs/opus/test/lapped_transform_unittest.cc b/third_party/libwebrtc/modules/audio_coding/codecs/opus/test/lapped_transform_unittest.cc new file mode 100644 index 0000000000..1003ed52e5 --- /dev/null +++ b/third_party/libwebrtc/modules/audio_coding/codecs/opus/test/lapped_transform_unittest.cc @@ -0,0 +1,203 @@ +/* + * Copyright (c) 2014 The WebRTC project authors. All Rights Reserved. + * + * Use of this source code is governed by a BSD-style license + * that can be found in the LICENSE file in the root of the source + * tree. An additional intellectual property rights grant can be found + * in the file PATENTS. All contributing project authors may + * be found in the AUTHORS file in the root of the source tree. + */ + +#include "modules/audio_coding/codecs/opus/test/lapped_transform.h" + +#include +#include +#include + +#include "test/gtest.h" + +using std::complex; + +namespace { + +class NoopCallback : public webrtc::LappedTransform::Callback { + public: + NoopCallback() : block_num_(0) {} + + void ProcessAudioBlock(const complex* const* in_block, + size_t in_channels, + size_t frames, + size_t out_channels, + complex* const* out_block) override { + RTC_CHECK_EQ(in_channels, out_channels); + for (size_t i = 0; i < out_channels; ++i) { + memcpy(out_block[i], in_block[i], sizeof(**in_block) * frames); + } + ++block_num_; + } + + size_t block_num() { return block_num_; } + + private: + size_t block_num_; +}; + +class FftCheckerCallback : public webrtc::LappedTransform::Callback { + public: + FftCheckerCallback() : block_num_(0) {} + + void ProcessAudioBlock(const complex* const* in_block, + size_t in_channels, + size_t frames, + size_t out_channels, + complex* const* out_block) override { + RTC_CHECK_EQ(in_channels, out_channels); + + size_t full_length = (frames - 1) * 2; + ++block_num_; + + if (block_num_ > 0) { + ASSERT_NEAR(in_block[0][0].real(), static_cast(full_length), + 1e-5f); + ASSERT_NEAR(in_block[0][0].imag(), 0.0f, 1e-5f); + for (size_t i = 1; i < frames; ++i) { + ASSERT_NEAR(in_block[0][i].real(), 0.0f, 1e-5f); + ASSERT_NEAR(in_block[0][i].imag(), 0.0f, 1e-5f); + } + } + } + + size_t block_num() { return block_num_; } + + private: + size_t block_num_; +}; + +void SetFloatArray(float value, int rows, int cols, float* const* array) { + for (int i = 0; i < rows; ++i) { + for (int j = 0; j < cols; ++j) { + array[i][j] = value; + } + } +} + +} // namespace + +namespace webrtc { + +TEST(LappedTransformTest, Windowless) { + const size_t kChannels = 3; + const size_t kChunkLength = 512; + const size_t kBlockLength = 64; + const size_t kShiftAmount = 64; + NoopCallback noop; + + // Rectangular window. + float window[kBlockLength]; + std::fill(window, &window[kBlockLength], 1.0f); + + LappedTransform trans(kChannels, kChannels, kChunkLength, window, + kBlockLength, kShiftAmount, &noop); + float in_buffer[kChannels][kChunkLength]; + float* in_chunk[kChannels]; + float out_buffer[kChannels][kChunkLength]; + float* out_chunk[kChannels]; + + in_chunk[0] = in_buffer[0]; + in_chunk[1] = in_buffer[1]; + in_chunk[2] = in_buffer[2]; + out_chunk[0] = out_buffer[0]; + out_chunk[1] = out_buffer[1]; + out_chunk[2] = out_buffer[2]; + SetFloatArray(2.0f, kChannels, kChunkLength, in_chunk); + SetFloatArray(-1.0f, kChannels, kChunkLength, out_chunk); + + trans.ProcessChunk(in_chunk, out_chunk); + + for (size_t i = 0; i < kChannels; ++i) { + for (size_t j = 0; j < kChunkLength; ++j) { + ASSERT_NEAR(out_chunk[i][j], 2.0f, 1e-5f); + } + } + + ASSERT_EQ(kChunkLength / kBlockLength, noop.block_num()); +} + +TEST(LappedTransformTest, IdentityProcessor) { + const size_t kChunkLength = 512; + const size_t kBlockLength = 64; + const size_t kShiftAmount = 32; + NoopCallback noop; + + // Identity window for |overlap = block_size / 2|. + float window[kBlockLength]; + std::fill(window, &window[kBlockLength], std::sqrt(0.5f)); + + LappedTransform trans(1, 1, kChunkLength, window, kBlockLength, kShiftAmount, + &noop); + float in_buffer[kChunkLength]; + float* in_chunk = in_buffer; + float out_buffer[kChunkLength]; + float* out_chunk = out_buffer; + + SetFloatArray(2.0f, 1, kChunkLength, &in_chunk); + SetFloatArray(-1.0f, 1, kChunkLength, &out_chunk); + + trans.ProcessChunk(&in_chunk, &out_chunk); + + for (size_t i = 0; i < kChunkLength; ++i) { + ASSERT_NEAR(out_chunk[i], (i < kBlockLength - kShiftAmount) ? 0.0f : 2.0f, + 1e-5f); + } + + ASSERT_EQ(kChunkLength / kShiftAmount, noop.block_num()); +} + +TEST(LappedTransformTest, Callbacks) { + const size_t kChunkLength = 512; + const size_t kBlockLength = 64; + FftCheckerCallback call; + + // Rectangular window. + float window[kBlockLength]; + std::fill(window, &window[kBlockLength], 1.0f); + + LappedTransform trans(1, 1, kChunkLength, window, kBlockLength, kBlockLength, + &call); + float in_buffer[kChunkLength]; + float* in_chunk = in_buffer; + float out_buffer[kChunkLength]; + float* out_chunk = out_buffer; + + SetFloatArray(1.0f, 1, kChunkLength, &in_chunk); + SetFloatArray(-1.0f, 1, kChunkLength, &out_chunk); + + trans.ProcessChunk(&in_chunk, &out_chunk); + + ASSERT_EQ(kChunkLength / kBlockLength, call.block_num()); +} + +TEST(LappedTransformTest, chunk_length) { + const size_t kBlockLength = 64; + FftCheckerCallback call; + const float window[kBlockLength] = {}; + + // Make sure that chunk_length returns the same value passed to the + // LappedTransform constructor. + { + const size_t kExpectedChunkLength = 512; + const LappedTransform trans(1, 1, kExpectedChunkLength, window, + kBlockLength, kBlockLength, &call); + + EXPECT_EQ(kExpectedChunkLength, trans.chunk_length()); + } + { + const size_t kExpectedChunkLength = 160; + const LappedTransform trans(1, 1, kExpectedChunkLength, window, + kBlockLength, kBlockLength, &call); + + EXPECT_EQ(kExpectedChunkLength, trans.chunk_length()); + } +} + +} // namespace webrtc -- cgit v1.2.3