webrtcvad 0.1.0 → 0.2.3
This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
- checksums.yaml +4 -4
- data/ext/webrtcvad/extconf.rb +29 -0
- data/ext/webrtcvad/webrtc/common_audio/signal_processing/division_operations.c +141 -0
- data/ext/webrtcvad/webrtc/common_audio/signal_processing/dot_product_with_scale.h +40 -0
- data/ext/webrtcvad/webrtc/common_audio/signal_processing/energy.c +39 -0
- data/ext/webrtcvad/webrtc/common_audio/signal_processing/get_scaling_square.c +46 -0
- data/ext/webrtcvad/webrtc/common_audio/signal_processing/include/signal_processing_library.h +1605 -0
- data/ext/webrtcvad/webrtc/common_audio/signal_processing/include/spl_inl.h +153 -0
- data/ext/webrtcvad/webrtc/common_audio/signal_processing/resample_48khz.c +186 -0
- data/ext/webrtcvad/webrtc/common_audio/signal_processing/resample_by_2_internal.c +689 -0
- data/ext/webrtcvad/webrtc/common_audio/signal_processing/resample_by_2_internal.h +60 -0
- data/ext/webrtcvad/webrtc/common_audio/signal_processing/resample_fractional.c +239 -0
- data/ext/webrtcvad/webrtc/common_audio/third_party/spl_sqrt_floor/spl_sqrt_floor.c +77 -0
- data/ext/webrtcvad/webrtc/common_audio/third_party/spl_sqrt_floor/spl_sqrt_floor.h +29 -0
- data/ext/webrtcvad/webrtc/common_audio/third_party/spl_sqrt_floor/spl_sqrt_floor_mips.c +207 -0
- data/ext/webrtcvad/webrtc/common_audio/vad/include/webrtc_vad.h +87 -0
- data/ext/webrtcvad/webrtc/common_audio/vad/vad_core.c +685 -0
- data/ext/webrtcvad/webrtc/common_audio/vad/vad_core.h +114 -0
- data/ext/webrtcvad/webrtc/common_audio/vad/vad_filterbank.c +329 -0
- data/ext/webrtcvad/webrtc/common_audio/vad/vad_filterbank.h +45 -0
- data/ext/webrtcvad/webrtc/common_audio/vad/vad_gmm.c +82 -0
- data/ext/webrtcvad/webrtc/common_audio/vad/vad_gmm.h +39 -0
- data/ext/webrtcvad/webrtc/common_audio/vad/vad_sp.c +176 -0
- data/ext/webrtcvad/webrtc/common_audio/vad/vad_sp.h +54 -0
- data/ext/webrtcvad/webrtc/common_audio/vad/webrtc_vad.c +114 -0
- data/ext/webrtcvad/webrtc/rtc_base/checks.cc +207 -0
- data/ext/webrtcvad/webrtc/rtc_base/checks.h +400 -0
- data/ext/webrtcvad/webrtc/rtc_base/compile_assert_c.h +25 -0
- data/ext/webrtcvad/webrtc/rtc_base/numerics/safe_compare.h +176 -0
- data/ext/webrtcvad/webrtc/rtc_base/sanitizer.h +144 -0
- data/ext/webrtcvad/webrtc/rtc_base/system/inline.h +31 -0
- data/ext/webrtcvad/webrtc/rtc_base/system/rtc_export.h +43 -0
- data/ext/webrtcvad/webrtc/rtc_base/type_traits.h +140 -0
- data/ext/webrtcvad/webrtcvad.c +112 -0
- metadata +37 -3
@@ -0,0 +1,82 @@
|
|
1
|
+
/*
|
2
|
+
* Copyright (c) 2011 The WebRTC project authors. All Rights Reserved.
|
3
|
+
*
|
4
|
+
* Use of this source code is governed by a BSD-style license
|
5
|
+
* that can be found in the LICENSE file in the root of the source
|
6
|
+
* tree. An additional intellectual property rights grant can be found
|
7
|
+
* in the file PATENTS. All contributing project authors may
|
8
|
+
* be found in the AUTHORS file in the root of the source tree.
|
9
|
+
*/
|
10
|
+
|
11
|
+
#include "common_audio/vad/vad_gmm.h"
|
12
|
+
|
13
|
+
#include "common_audio/signal_processing/include/signal_processing_library.h"
|
14
|
+
|
15
|
+
static const int32_t kCompVar = 22005;
|
16
|
+
static const int16_t kLog2Exp = 5909; // log2(exp(1)) in Q12.
|
17
|
+
|
18
|
+
// For a normal distribution, the probability of |input| is calculated and
|
19
|
+
// returned (in Q20). The formula for normal distributed probability is
|
20
|
+
//
|
21
|
+
// 1 / s * exp(-(x - m)^2 / (2 * s^2))
|
22
|
+
//
|
23
|
+
// where the parameters are given in the following Q domains:
|
24
|
+
// m = |mean| (Q7)
|
25
|
+
// s = |std| (Q7)
|
26
|
+
// x = |input| (Q4)
|
27
|
+
// in addition to the probability we output |delta| (in Q11) used when updating
|
28
|
+
// the noise/speech model.
|
29
|
+
int32_t WebRtcVad_GaussianProbability(int16_t input,
|
30
|
+
int16_t mean,
|
31
|
+
int16_t std,
|
32
|
+
int16_t* delta) {
|
33
|
+
int16_t tmp16, inv_std, inv_std2, exp_value = 0;
|
34
|
+
int32_t tmp32;
|
35
|
+
|
36
|
+
// Calculate |inv_std| = 1 / s, in Q10.
|
37
|
+
// 131072 = 1 in Q17, and (|std| >> 1) is for rounding instead of truncation.
|
38
|
+
// Q-domain: Q17 / Q7 = Q10.
|
39
|
+
tmp32 = (int32_t) 131072 + (int32_t) (std >> 1);
|
40
|
+
inv_std = (int16_t) WebRtcSpl_DivW32W16(tmp32, std);
|
41
|
+
|
42
|
+
// Calculate |inv_std2| = 1 / s^2, in Q14.
|
43
|
+
tmp16 = (inv_std >> 2); // Q10 -> Q8.
|
44
|
+
// Q-domain: (Q8 * Q8) >> 2 = Q14.
|
45
|
+
inv_std2 = (int16_t)((tmp16 * tmp16) >> 2);
|
46
|
+
// TODO(bjornv): Investigate if changing to
|
47
|
+
// inv_std2 = (int16_t)((inv_std * inv_std) >> 6);
|
48
|
+
// gives better accuracy.
|
49
|
+
|
50
|
+
tmp16 = (input << 3); // Q4 -> Q7
|
51
|
+
tmp16 = tmp16 - mean; // Q7 - Q7 = Q7
|
52
|
+
|
53
|
+
// To be used later, when updating noise/speech model.
|
54
|
+
// |delta| = (x - m) / s^2, in Q11.
|
55
|
+
// Q-domain: (Q14 * Q7) >> 10 = Q11.
|
56
|
+
*delta = (int16_t)((inv_std2 * tmp16) >> 10);
|
57
|
+
|
58
|
+
// Calculate the exponent |tmp32| = (x - m)^2 / (2 * s^2), in Q10. Replacing
|
59
|
+
// division by two with one shift.
|
60
|
+
// Q-domain: (Q11 * Q7) >> 8 = Q10.
|
61
|
+
tmp32 = (*delta * tmp16) >> 9;
|
62
|
+
|
63
|
+
// If the exponent is small enough to give a non-zero probability we calculate
|
64
|
+
// |exp_value| ~= exp(-(x - m)^2 / (2 * s^2))
|
65
|
+
// ~= exp2(-log2(exp(1)) * |tmp32|).
|
66
|
+
if (tmp32 < kCompVar) {
|
67
|
+
// Calculate |tmp16| = log2(exp(1)) * |tmp32|, in Q10.
|
68
|
+
// Q-domain: (Q12 * Q10) >> 12 = Q10.
|
69
|
+
tmp16 = (int16_t)((kLog2Exp * tmp32) >> 12);
|
70
|
+
tmp16 = -tmp16;
|
71
|
+
exp_value = (0x0400 | (tmp16 & 0x03FF));
|
72
|
+
tmp16 ^= 0xFFFF;
|
73
|
+
tmp16 >>= 10;
|
74
|
+
tmp16 += 1;
|
75
|
+
// Get |exp_value| = exp(-|tmp32|) in Q10.
|
76
|
+
exp_value >>= tmp16;
|
77
|
+
}
|
78
|
+
|
79
|
+
// Calculate and return (1 / s) * exp(-(x - m)^2 / (2 * s^2)), in Q20.
|
80
|
+
// Q-domain: Q10 * Q10 = Q20.
|
81
|
+
return inv_std * exp_value;
|
82
|
+
}
|
@@ -0,0 +1,39 @@
|
|
1
|
+
/*
|
2
|
+
* Copyright (c) 2011 The WebRTC project authors. All Rights Reserved.
|
3
|
+
*
|
4
|
+
* Use of this source code is governed by a BSD-style license
|
5
|
+
* that can be found in the LICENSE file in the root of the source
|
6
|
+
* tree. An additional intellectual property rights grant can be found
|
7
|
+
* in the file PATENTS. All contributing project authors may
|
8
|
+
* be found in the AUTHORS file in the root of the source tree.
|
9
|
+
*/
|
10
|
+
|
11
|
+
// Gaussian probability calculations internally used in vad_core.c.
|
12
|
+
|
13
|
+
#ifndef COMMON_AUDIO_VAD_VAD_GMM_H_
|
14
|
+
#define COMMON_AUDIO_VAD_VAD_GMM_H_
|
15
|
+
|
16
|
+
#include <stdint.h>
|
17
|
+
|
18
|
+
// Calculates the probability for |input|, given that |input| comes from a
|
19
|
+
// normal distribution with mean and standard deviation (|mean|, |std|).
|
20
|
+
//
|
21
|
+
// Inputs:
|
22
|
+
// - input : input sample in Q4.
|
23
|
+
// - mean : mean input in the statistical model, Q7.
|
24
|
+
// - std : standard deviation, Q7.
|
25
|
+
//
|
26
|
+
// Output:
|
27
|
+
//
|
28
|
+
// - delta : input used when updating the model, Q11.
|
29
|
+
// |delta| = (|input| - |mean|) / |std|^2.
|
30
|
+
//
|
31
|
+
// Return:
|
32
|
+
// (probability for |input|) =
|
33
|
+
// 1 / |std| * exp(-(|input| - |mean|)^2 / (2 * |std|^2));
|
34
|
+
int32_t WebRtcVad_GaussianProbability(int16_t input,
|
35
|
+
int16_t mean,
|
36
|
+
int16_t std,
|
37
|
+
int16_t* delta);
|
38
|
+
|
39
|
+
#endif // COMMON_AUDIO_VAD_VAD_GMM_H_
|
@@ -0,0 +1,176 @@
|
|
1
|
+
/*
|
2
|
+
* Copyright (c) 2012 The WebRTC project authors. All Rights Reserved.
|
3
|
+
*
|
4
|
+
* Use of this source code is governed by a BSD-style license
|
5
|
+
* that can be found in the LICENSE file in the root of the source
|
6
|
+
* tree. An additional intellectual property rights grant can be found
|
7
|
+
* in the file PATENTS. All contributing project authors may
|
8
|
+
* be found in the AUTHORS file in the root of the source tree.
|
9
|
+
*/
|
10
|
+
|
11
|
+
#include "common_audio/vad/vad_sp.h"
|
12
|
+
|
13
|
+
#include "rtc_base/checks.h"
|
14
|
+
#include "common_audio/signal_processing/include/signal_processing_library.h"
|
15
|
+
#include "common_audio/vad/vad_core.h"
|
16
|
+
|
17
|
+
// Allpass filter coefficients, upper and lower, in Q13.
|
18
|
+
// Upper: 0.64, Lower: 0.17.
|
19
|
+
static const int16_t kAllPassCoefsQ13[2] = { 5243, 1392 }; // Q13.
|
20
|
+
static const int16_t kSmoothingDown = 6553; // 0.2 in Q15.
|
21
|
+
static const int16_t kSmoothingUp = 32439; // 0.99 in Q15.
|
22
|
+
|
23
|
+
// TODO(bjornv): Move this function to vad_filterbank.c.
|
24
|
+
// Downsampling filter based on splitting filter and allpass functions.
|
25
|
+
void WebRtcVad_Downsampling(const int16_t* signal_in,
|
26
|
+
int16_t* signal_out,
|
27
|
+
int32_t* filter_state,
|
28
|
+
size_t in_length) {
|
29
|
+
int16_t tmp16_1 = 0, tmp16_2 = 0;
|
30
|
+
int32_t tmp32_1 = filter_state[0];
|
31
|
+
int32_t tmp32_2 = filter_state[1];
|
32
|
+
size_t n = 0;
|
33
|
+
// Downsampling by 2 gives half length.
|
34
|
+
size_t half_length = (in_length >> 1);
|
35
|
+
|
36
|
+
// Filter coefficients in Q13, filter state in Q0.
|
37
|
+
for (n = 0; n < half_length; n++) {
|
38
|
+
// All-pass filtering upper branch.
|
39
|
+
tmp16_1 = (int16_t) ((tmp32_1 >> 1) +
|
40
|
+
((kAllPassCoefsQ13[0] * *signal_in) >> 14));
|
41
|
+
*signal_out = tmp16_1;
|
42
|
+
tmp32_1 = (int32_t)(*signal_in++) - ((kAllPassCoefsQ13[0] * tmp16_1) >> 12);
|
43
|
+
|
44
|
+
// All-pass filtering lower branch.
|
45
|
+
tmp16_2 = (int16_t) ((tmp32_2 >> 1) +
|
46
|
+
((kAllPassCoefsQ13[1] * *signal_in) >> 14));
|
47
|
+
*signal_out++ += tmp16_2;
|
48
|
+
tmp32_2 = (int32_t)(*signal_in++) - ((kAllPassCoefsQ13[1] * tmp16_2) >> 12);
|
49
|
+
}
|
50
|
+
// Store the filter states.
|
51
|
+
filter_state[0] = tmp32_1;
|
52
|
+
filter_state[1] = tmp32_2;
|
53
|
+
}
|
54
|
+
|
55
|
+
// Inserts |feature_value| into |low_value_vector|, if it is one of the 16
|
56
|
+
// smallest values the last 100 frames. Then calculates and returns the median
|
57
|
+
// of the five smallest values.
|
58
|
+
int16_t WebRtcVad_FindMinimum(VadInstT* self,
|
59
|
+
int16_t feature_value,
|
60
|
+
int channel) {
|
61
|
+
int i = 0, j = 0;
|
62
|
+
int position = -1;
|
63
|
+
// Offset to beginning of the 16 minimum values in memory.
|
64
|
+
const int offset = (channel << 4);
|
65
|
+
int16_t current_median = 1600;
|
66
|
+
int16_t alpha = 0;
|
67
|
+
int32_t tmp32 = 0;
|
68
|
+
// Pointer to memory for the 16 minimum values and the age of each value of
|
69
|
+
// the |channel|.
|
70
|
+
int16_t* age = &self->index_vector[offset];
|
71
|
+
int16_t* smallest_values = &self->low_value_vector[offset];
|
72
|
+
|
73
|
+
RTC_DCHECK_LT(channel, kNumChannels);
|
74
|
+
|
75
|
+
// Each value in |smallest_values| is getting 1 loop older. Update |age|, and
|
76
|
+
// remove old values.
|
77
|
+
for (i = 0; i < 16; i++) {
|
78
|
+
if (age[i] != 100) {
|
79
|
+
age[i]++;
|
80
|
+
} else {
|
81
|
+
// Too old value. Remove from memory and shift larger values downwards.
|
82
|
+
for (j = i; j < 15; j++) {
|
83
|
+
smallest_values[j] = smallest_values[j + 1];
|
84
|
+
age[j] = age[j + 1];
|
85
|
+
}
|
86
|
+
age[15] = 101;
|
87
|
+
smallest_values[15] = 10000;
|
88
|
+
}
|
89
|
+
}
|
90
|
+
|
91
|
+
// Check if |feature_value| is smaller than any of the values in
|
92
|
+
// |smallest_values|. If so, find the |position| where to insert the new value
|
93
|
+
// (|feature_value|).
|
94
|
+
if (feature_value < smallest_values[7]) {
|
95
|
+
if (feature_value < smallest_values[3]) {
|
96
|
+
if (feature_value < smallest_values[1]) {
|
97
|
+
if (feature_value < smallest_values[0]) {
|
98
|
+
position = 0;
|
99
|
+
} else {
|
100
|
+
position = 1;
|
101
|
+
}
|
102
|
+
} else if (feature_value < smallest_values[2]) {
|
103
|
+
position = 2;
|
104
|
+
} else {
|
105
|
+
position = 3;
|
106
|
+
}
|
107
|
+
} else if (feature_value < smallest_values[5]) {
|
108
|
+
if (feature_value < smallest_values[4]) {
|
109
|
+
position = 4;
|
110
|
+
} else {
|
111
|
+
position = 5;
|
112
|
+
}
|
113
|
+
} else if (feature_value < smallest_values[6]) {
|
114
|
+
position = 6;
|
115
|
+
} else {
|
116
|
+
position = 7;
|
117
|
+
}
|
118
|
+
} else if (feature_value < smallest_values[15]) {
|
119
|
+
if (feature_value < smallest_values[11]) {
|
120
|
+
if (feature_value < smallest_values[9]) {
|
121
|
+
if (feature_value < smallest_values[8]) {
|
122
|
+
position = 8;
|
123
|
+
} else {
|
124
|
+
position = 9;
|
125
|
+
}
|
126
|
+
} else if (feature_value < smallest_values[10]) {
|
127
|
+
position = 10;
|
128
|
+
} else {
|
129
|
+
position = 11;
|
130
|
+
}
|
131
|
+
} else if (feature_value < smallest_values[13]) {
|
132
|
+
if (feature_value < smallest_values[12]) {
|
133
|
+
position = 12;
|
134
|
+
} else {
|
135
|
+
position = 13;
|
136
|
+
}
|
137
|
+
} else if (feature_value < smallest_values[14]) {
|
138
|
+
position = 14;
|
139
|
+
} else {
|
140
|
+
position = 15;
|
141
|
+
}
|
142
|
+
}
|
143
|
+
|
144
|
+
// If we have detected a new small value, insert it at the correct position
|
145
|
+
// and shift larger values up.
|
146
|
+
if (position > -1) {
|
147
|
+
for (i = 15; i > position; i--) {
|
148
|
+
smallest_values[i] = smallest_values[i - 1];
|
149
|
+
age[i] = age[i - 1];
|
150
|
+
}
|
151
|
+
smallest_values[position] = feature_value;
|
152
|
+
age[position] = 1;
|
153
|
+
}
|
154
|
+
|
155
|
+
// Get |current_median|.
|
156
|
+
if (self->frame_counter > 2) {
|
157
|
+
current_median = smallest_values[2];
|
158
|
+
} else if (self->frame_counter > 0) {
|
159
|
+
current_median = smallest_values[0];
|
160
|
+
}
|
161
|
+
|
162
|
+
// Smooth the median value.
|
163
|
+
if (self->frame_counter > 0) {
|
164
|
+
if (current_median < self->mean_value[channel]) {
|
165
|
+
alpha = kSmoothingDown; // 0.2 in Q15.
|
166
|
+
} else {
|
167
|
+
alpha = kSmoothingUp; // 0.99 in Q15.
|
168
|
+
}
|
169
|
+
}
|
170
|
+
tmp32 = (alpha + 1) * self->mean_value[channel];
|
171
|
+
tmp32 += (WEBRTC_SPL_WORD16_MAX - alpha) * current_median;
|
172
|
+
tmp32 += 16384;
|
173
|
+
self->mean_value[channel] = (int16_t) (tmp32 >> 15);
|
174
|
+
|
175
|
+
return self->mean_value[channel];
|
176
|
+
}
|
@@ -0,0 +1,54 @@
|
|
1
|
+
/*
|
2
|
+
* Copyright (c) 2011 The WebRTC project authors. All Rights Reserved.
|
3
|
+
*
|
4
|
+
* Use of this source code is governed by a BSD-style license
|
5
|
+
* that can be found in the LICENSE file in the root of the source
|
6
|
+
* tree. An additional intellectual property rights grant can be found
|
7
|
+
* in the file PATENTS. All contributing project authors may
|
8
|
+
* be found in the AUTHORS file in the root of the source tree.
|
9
|
+
*/
|
10
|
+
|
11
|
+
// This file includes specific signal processing tools used in vad_core.c.
|
12
|
+
|
13
|
+
#ifndef COMMON_AUDIO_VAD_VAD_SP_H_
|
14
|
+
#define COMMON_AUDIO_VAD_VAD_SP_H_
|
15
|
+
|
16
|
+
#include "common_audio/vad/vad_core.h"
|
17
|
+
|
18
|
+
// Downsamples the signal by a factor 2, eg. 32->16 or 16->8.
|
19
|
+
//
|
20
|
+
// Inputs:
|
21
|
+
// - signal_in : Input signal.
|
22
|
+
// - in_length : Length of input signal in samples.
|
23
|
+
//
|
24
|
+
// Input & Output:
|
25
|
+
// - filter_state : Current filter states of the two all-pass filters. The
|
26
|
+
// |filter_state| is updated after all samples have been
|
27
|
+
// processed.
|
28
|
+
//
|
29
|
+
// Output:
|
30
|
+
// - signal_out : Downsampled signal (of length |in_length| / 2).
|
31
|
+
void WebRtcVad_Downsampling(const int16_t* signal_in,
|
32
|
+
int16_t* signal_out,
|
33
|
+
int32_t* filter_state,
|
34
|
+
size_t in_length);
|
35
|
+
|
36
|
+
// Updates and returns the smoothed feature minimum. As minimum we use the
|
37
|
+
// median of the five smallest feature values in a 100 frames long window.
|
38
|
+
// As long as |handle->frame_counter| is zero, that is, we haven't received any
|
39
|
+
// "valid" data, FindMinimum() outputs the default value of 1600.
|
40
|
+
//
|
41
|
+
// Inputs:
|
42
|
+
// - feature_value : New feature value to update with.
|
43
|
+
// - channel : Channel number.
|
44
|
+
//
|
45
|
+
// Input & Output:
|
46
|
+
// - handle : State information of the VAD.
|
47
|
+
//
|
48
|
+
// Returns:
|
49
|
+
// : Smoothed minimum value for a moving window.
|
50
|
+
int16_t WebRtcVad_FindMinimum(VadInstT* handle,
|
51
|
+
int16_t feature_value,
|
52
|
+
int channel);
|
53
|
+
|
54
|
+
#endif // COMMON_AUDIO_VAD_VAD_SP_H_
|
@@ -0,0 +1,114 @@
|
|
1
|
+
/*
|
2
|
+
* Copyright (c) 2012 The WebRTC project authors. All Rights Reserved.
|
3
|
+
*
|
4
|
+
* Use of this source code is governed by a BSD-style license
|
5
|
+
* that can be found in the LICENSE file in the root of the source
|
6
|
+
* tree. An additional intellectual property rights grant can be found
|
7
|
+
* in the file PATENTS. All contributing project authors may
|
8
|
+
* be found in the AUTHORS file in the root of the source tree.
|
9
|
+
*/
|
10
|
+
|
11
|
+
#include "common_audio/vad/include/webrtc_vad.h"
|
12
|
+
|
13
|
+
#include <stdlib.h>
|
14
|
+
#include <string.h>
|
15
|
+
|
16
|
+
#include "common_audio/signal_processing/include/signal_processing_library.h"
|
17
|
+
#include "common_audio/vad/vad_core.h"
|
18
|
+
|
19
|
+
static const int kInitCheck = 42;
|
20
|
+
static const int kValidRates[] = { 8000, 16000, 32000, 48000 };
|
21
|
+
static const size_t kRatesSize = sizeof(kValidRates) / sizeof(*kValidRates);
|
22
|
+
static const int kMaxFrameLengthMs = 30;
|
23
|
+
|
24
|
+
VadInst* WebRtcVad_Create() {
|
25
|
+
VadInstT* self = (VadInstT*)malloc(sizeof(VadInstT));
|
26
|
+
|
27
|
+
self->init_flag = 0;
|
28
|
+
|
29
|
+
return (VadInst*)self;
|
30
|
+
}
|
31
|
+
|
32
|
+
void WebRtcVad_Free(VadInst* handle) {
|
33
|
+
free(handle);
|
34
|
+
}
|
35
|
+
|
36
|
+
// TODO(bjornv): Move WebRtcVad_InitCore() code here.
|
37
|
+
int WebRtcVad_Init(VadInst* handle) {
|
38
|
+
// Initialize the core VAD component.
|
39
|
+
return WebRtcVad_InitCore((VadInstT*) handle);
|
40
|
+
}
|
41
|
+
|
42
|
+
// TODO(bjornv): Move WebRtcVad_set_mode_core() code here.
|
43
|
+
int WebRtcVad_set_mode(VadInst* handle, int mode) {
|
44
|
+
VadInstT* self = (VadInstT*) handle;
|
45
|
+
|
46
|
+
if (handle == NULL) {
|
47
|
+
return -1;
|
48
|
+
}
|
49
|
+
if (self->init_flag != kInitCheck) {
|
50
|
+
return -1;
|
51
|
+
}
|
52
|
+
|
53
|
+
return WebRtcVad_set_mode_core(self, mode);
|
54
|
+
}
|
55
|
+
|
56
|
+
int WebRtcVad_Process(VadInst* handle, int fs, const int16_t* audio_frame,
|
57
|
+
size_t frame_length) {
|
58
|
+
int vad = -1;
|
59
|
+
VadInstT* self = (VadInstT*) handle;
|
60
|
+
|
61
|
+
if (handle == NULL) {
|
62
|
+
return -1;
|
63
|
+
}
|
64
|
+
|
65
|
+
if (self->init_flag != kInitCheck) {
|
66
|
+
return -1;
|
67
|
+
}
|
68
|
+
if (audio_frame == NULL) {
|
69
|
+
return -1;
|
70
|
+
}
|
71
|
+
if (WebRtcVad_ValidRateAndFrameLength(fs, frame_length) != 0) {
|
72
|
+
return -1;
|
73
|
+
}
|
74
|
+
|
75
|
+
if (fs == 48000) {
|
76
|
+
vad = WebRtcVad_CalcVad48khz(self, audio_frame, frame_length);
|
77
|
+
} else if (fs == 32000) {
|
78
|
+
vad = WebRtcVad_CalcVad32khz(self, audio_frame, frame_length);
|
79
|
+
} else if (fs == 16000) {
|
80
|
+
vad = WebRtcVad_CalcVad16khz(self, audio_frame, frame_length);
|
81
|
+
} else if (fs == 8000) {
|
82
|
+
vad = WebRtcVad_CalcVad8khz(self, audio_frame, frame_length);
|
83
|
+
}
|
84
|
+
|
85
|
+
if (vad > 0) {
|
86
|
+
vad = 1;
|
87
|
+
}
|
88
|
+
return vad;
|
89
|
+
}
|
90
|
+
|
91
|
+
int WebRtcVad_ValidRateAndFrameLength(int rate, size_t frame_length) {
|
92
|
+
int return_value = -1;
|
93
|
+
size_t i;
|
94
|
+
int valid_length_ms;
|
95
|
+
size_t valid_length;
|
96
|
+
|
97
|
+
// We only allow 10, 20 or 30 ms frames. Loop through valid frame rates and
|
98
|
+
// see if we have a matching pair.
|
99
|
+
for (i = 0; i < kRatesSize; i++) {
|
100
|
+
if (kValidRates[i] == rate) {
|
101
|
+
for (valid_length_ms = 10; valid_length_ms <= kMaxFrameLengthMs;
|
102
|
+
valid_length_ms += 10) {
|
103
|
+
valid_length = (size_t)(kValidRates[i] / 1000 * valid_length_ms);
|
104
|
+
if (frame_length == valid_length) {
|
105
|
+
return_value = 0;
|
106
|
+
break;
|
107
|
+
}
|
108
|
+
}
|
109
|
+
break;
|
110
|
+
}
|
111
|
+
}
|
112
|
+
|
113
|
+
return return_value;
|
114
|
+
}
|