1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
|
/*
* 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/neteq/accelerate.h"
#include "api/array_view.h"
#include "modules/audio_coding/neteq/audio_multi_vector.h"
namespace webrtc {
Accelerate::ReturnCodes Accelerate::Process(const int16_t* input,
size_t input_length,
bool fast_accelerate,
AudioMultiVector* output,
size_t* length_change_samples) {
// Input length must be (almost) 30 ms.
static const size_t k15ms = 120; // 15 ms = 120 samples at 8 kHz sample rate.
if (num_channels_ == 0 ||
input_length / num_channels_ < (2 * k15ms - 1) * fs_mult_) {
// Length of input data too short to do accelerate. Simply move all data
// from input to output.
output->PushBackInterleaved(
rtc::ArrayView<const int16_t>(input, input_length));
return kError;
}
return TimeStretch::Process(input, input_length, fast_accelerate, output,
length_change_samples);
}
void Accelerate::SetParametersForPassiveSpeech(size_t /*len*/,
int16_t* best_correlation,
size_t* /*peak_index*/) const {
// When the signal does not contain any active speech, the correlation does
// not matter. Simply set it to zero.
*best_correlation = 0;
}
Accelerate::ReturnCodes Accelerate::CheckCriteriaAndStretch(
const int16_t* input,
size_t input_length,
size_t peak_index,
int16_t best_correlation,
bool active_speech,
bool fast_mode,
AudioMultiVector* output) const {
// Check for strong correlation or passive speech.
// Use 8192 (0.5 in Q14) in fast mode.
const int correlation_threshold = fast_mode ? 8192 : kCorrelationThreshold;
if ((best_correlation > correlation_threshold) || !active_speech) {
// Do accelerate operation by overlap add.
// Pre-calculate common multiplication with `fs_mult_`.
// 120 corresponds to 15 ms.
size_t fs_mult_120 = fs_mult_ * 120;
if (fast_mode) {
// Fit as many multiples of `peak_index` as possible in fs_mult_120.
// TODO(henrik.lundin) Consider finding multiple correlation peaks and
// pick the one with the longest correlation lag in this case.
peak_index = (fs_mult_120 / peak_index) * peak_index;
}
RTC_DCHECK_GE(fs_mult_120, peak_index); // Should be handled in Process().
// Copy first part; 0 to 15 ms.
output->PushBackInterleaved(
rtc::ArrayView<const int16_t>(input, fs_mult_120 * num_channels_));
// Copy the `peak_index` starting at 15 ms to `temp_vector`.
AudioMultiVector temp_vector(num_channels_);
temp_vector.PushBackInterleaved(rtc::ArrayView<const int16_t>(
&input[fs_mult_120 * num_channels_], peak_index * num_channels_));
// Cross-fade `temp_vector` onto the end of `output`.
output->CrossFade(temp_vector, peak_index);
// Copy the last unmodified part, 15 ms + pitch period until the end.
output->PushBackInterleaved(rtc::ArrayView<const int16_t>(
&input[(fs_mult_120 + peak_index) * num_channels_],
input_length - (fs_mult_120 + peak_index) * num_channels_));
if (active_speech) {
return kSuccess;
} else {
return kSuccessLowEnergy;
}
} else {
// Accelerate not allowed. Simply move all data from decoded to outData.
output->PushBackInterleaved(
rtc::ArrayView<const int16_t>(input, input_length));
return kNoStretch;
}
}
Accelerate* AccelerateFactory::Create(
int sample_rate_hz,
size_t num_channels,
const BackgroundNoise& background_noise) const {
return new Accelerate(sample_rate_hz, num_channels, background_noise);
}
} // namespace webrtc
|
