OLD | NEW |
---|---|
(Empty) | |
1 /* | |
2 * Copyright (c) 2015 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 "webrtc/modules/audio_processing/vad/voice_activity_detector.h" | |
12 | |
13 #include "webrtc/base/checks.h" | |
14 | |
15 namespace webrtc { | |
16 namespace { | |
17 | |
18 const int kMaxLength = 320; | |
19 const int kNumChannels = 1; | |
20 | |
21 const double kDefaultVoiceValue = 1.0; | |
22 const double kNeutralProbability = 0.5; | |
23 const double kLowProbability = 0.01; | |
24 | |
25 } // namespace | |
26 | |
27 VoiceActivityDetector::VoiceActivityDetector() | |
28 : last_voice_probability_(kDefaultVoiceValue), | |
29 // Initialize to the most common resampling situation. | |
30 resampler_(kMaxLength, kLength10Ms, kNumChannels), | |
31 standalone_vad_(StandaloneVad::Create()) { | |
32 } | |
33 | |
34 // Because ISAC has a different chunk length, it updates | |
35 // |chunkwise_voice_probabilities_| and |chunkwise_rms_| when there is new data. | |
36 // Else it clears them. | |
Andrew MacDonald
2015/06/16 22:18:32
Else -> Otherwise
aluebs-webrtc
2015/06/17 01:44:31
Done.
| |
37 void VoiceActivityDetector::ProcessChunk(const int16_t* audio, | |
38 int length, | |
39 int sample_rate_hz) { | |
40 DCHECK_EQ(length, sample_rate_hz / 100); | |
41 DCHECK_LE(length, kMaxLength); | |
42 // Resample to the required rate. | |
43 const int16_t* resampled_ptr = audio; | |
44 if (sample_rate_hz != kSampleRateHz) { | |
45 CHECK_EQ( | |
46 resampler_.ResetIfNeeded(sample_rate_hz, kSampleRateHz, kNumChannels), | |
47 0); | |
48 resampler_.Push(audio, length, resampled_, kLength10Ms, length); | |
49 resampled_ptr = resampled_; | |
50 } | |
51 DCHECK_EQ(length, kLength10Ms); | |
52 | |
53 CHECK_EQ(standalone_vad_->AddAudio(audio, length), 0); | |
54 | |
55 audio_processing_.ExtractFeatures(resampled_ptr, length, &features_); | |
56 | |
57 chunkwise_voice_probabilities_.clear(); | |
58 chunkwise_rms_.clear(); | |
59 if (features_.num_frames > 0) { | |
60 if (features_.silence) { | |
61 // The other features are invalid, so set the voice probabilities to an | |
62 // arbitrary low value. | |
63 chunkwise_voice_probabilities_.assign(features_.num_frames, | |
64 kLowProbability); | |
65 } else { | |
66 chunkwise_voice_probabilities_.assign(features_.num_frames, | |
67 kNeutralProbability); | |
68 CHECK_GE( | |
69 standalone_vad_->GetActivity(&chunkwise_voice_probabilities_[0], | |
70 chunkwise_voice_probabilities_.size()), | |
71 0); | |
72 CHECK_GE(pitch_based_vad_.VoicingProbability( | |
73 features_, &chunkwise_voice_probabilities_[0]), | |
74 0); | |
75 } | |
76 last_voice_probability_ = chunkwise_voice_probabilities_.back(); | |
77 } | |
78 for (int i = 0; i < features_.num_frames; ++i) { | |
79 chunkwise_rms_.push_back(features_.rms[i]); | |
80 } | |
81 } | |
82 | |
83 } // namespace webrtc | |
OLD | NEW |