OLD | NEW |
1 /* | 1 /* |
2 * Copyright (c) 2016 The WebRTC project authors. All Rights Reserved. | 2 * Copyright (c) 2016 The WebRTC project authors. All Rights Reserved. |
3 * | 3 * |
4 * Use of this source code is governed by a BSD-style license | 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 | 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 | 6 * tree. An additional intellectual property rights grant can be found |
7 * in the file PATENTS. All contributing project authors may | 7 * in the file PATENTS. All contributing project authors may |
8 * be found in the AUTHORS file in the root of the source tree. | 8 * be found in the AUTHORS file in the root of the source tree. |
9 */ | 9 */ |
10 | 10 |
11 #include "webrtc/modules/audio_processing/residual_echo_detector.h" | 11 #include "webrtc/modules/audio_processing/residual_echo_detector.h" |
12 | 12 |
| 13 #include <algorithm> |
| 14 #include <numeric> |
| 15 |
13 #include "webrtc/modules/audio_processing/audio_buffer.h" | 16 #include "webrtc/modules/audio_processing/audio_buffer.h" |
14 | 17 |
| 18 namespace { |
| 19 |
| 20 float Power(rtc::ArrayView<const float> input) { |
| 21 return std::inner_product(input.begin(), input.end(), input.begin(), 0.f); |
| 22 } |
| 23 |
| 24 constexpr size_t kLookbackFrames = 650; |
| 25 // TODO(ivoc): Verify the size of this buffer. |
| 26 constexpr size_t kRenderBufferSize = 30; |
| 27 |
| 28 } // namespace |
| 29 |
15 namespace webrtc { | 30 namespace webrtc { |
16 | 31 |
17 ResidualEchoDetector::ResidualEchoDetector() = default; | 32 ResidualEchoDetector::ResidualEchoDetector() |
| 33 : render_buffer_(kRenderBufferSize), |
| 34 render_power_(kLookbackFrames), |
| 35 render_power_mean_(kLookbackFrames), |
| 36 render_power_std_dev_(kLookbackFrames), |
| 37 covariances_(kLookbackFrames){}; |
18 | 38 |
19 ResidualEchoDetector::~ResidualEchoDetector() = default; | 39 ResidualEchoDetector::~ResidualEchoDetector() = default; |
20 | 40 |
21 void ResidualEchoDetector::AnalyzeRenderAudio( | 41 void ResidualEchoDetector::AnalyzeRenderAudio( |
22 rtc::ArrayView<const float> render_audio) const { | 42 rtc::ArrayView<const float> render_audio) { |
23 // TODO(ivoc): Add implementation. | 43 if (render_buffer_.Size() == 0) { |
| 44 frames_since_zero_buffer_size_ = 0; |
| 45 } else if (frames_since_zero_buffer_size_ >= kRenderBufferSize) { |
| 46 // This can happen in a few cases: at the start of a call, due to a glitch |
| 47 // or due to clock drift. The excess capture value will be ignored. |
| 48 // TODO(ivoc): Include how often this happens in APM stats. |
| 49 render_buffer_.Pop(); |
| 50 frames_since_zero_buffer_size_ = 0; |
| 51 } |
| 52 ++frames_since_zero_buffer_size_; |
| 53 float power = Power(render_audio); |
| 54 render_buffer_.Push(power); |
24 } | 55 } |
25 | 56 |
26 void ResidualEchoDetector::AnalyzeCaptureAudio( | 57 void ResidualEchoDetector::AnalyzeCaptureAudio( |
27 rtc::ArrayView<const float> render_audio) { | 58 rtc::ArrayView<const float> capture_audio) { |
28 // TODO(ivoc): Add implementation. | 59 if (first_process_call_) { |
| 60 // On the first process call (so the start of a call), we must flush the |
| 61 // render buffer, otherwise the render data will be delayed. |
| 62 render_buffer_.Clear(); |
| 63 first_process_call_ = false; |
| 64 } |
| 65 |
| 66 // Get the next render value. |
| 67 const rtc::Optional<float> buffered_render_power = render_buffer_.Pop(); |
| 68 if (!buffered_render_power) { |
| 69 // This can happen in a few cases: at the start of a call, due to a glitch |
| 70 // or due to clock drift. The excess capture value will be ignored. |
| 71 // TODO(ivoc): Include how often this happens in APM stats. |
| 72 return; |
| 73 } |
| 74 // Update the render statistics, and store the statistics in circular buffers. |
| 75 render_statistics_.Update(*buffered_render_power); |
| 76 RTC_DCHECK_LT(next_insertion_index_, kLookbackFrames); |
| 77 render_power_[next_insertion_index_] = *buffered_render_power; |
| 78 render_power_mean_[next_insertion_index_] = render_statistics_.mean(); |
| 79 render_power_std_dev_[next_insertion_index_] = |
| 80 render_statistics_.std_deviation(); |
| 81 |
| 82 // Get the next capture value, update capture statistics and add the relevant |
| 83 // values to the buffers. |
| 84 const float capture_power = Power(capture_audio); |
| 85 capture_statistics_.Update(capture_power); |
| 86 const float capture_mean = capture_statistics_.mean(); |
| 87 const float capture_std_deviation = capture_statistics_.std_deviation(); |
| 88 |
| 89 // Update the covariance values and determine the new echo likelihood. |
| 90 echo_likelihood_ = 0.f; |
| 91 for (size_t delay = 0; delay < covariances_.size(); ++delay) { |
| 92 const size_t read_index = |
| 93 (kLookbackFrames + next_insertion_index_ - delay) % kLookbackFrames; |
| 94 RTC_DCHECK_LT(read_index, render_power_.size()); |
| 95 covariances_[delay].Update(capture_power, capture_mean, |
| 96 capture_std_deviation, render_power_[read_index], |
| 97 render_power_mean_[read_index], |
| 98 render_power_std_dev_[read_index]); |
| 99 echo_likelihood_ = std::max( |
| 100 echo_likelihood_, covariances_[delay].normalized_cross_correlation()); |
| 101 } |
| 102 |
| 103 // Update the next insertion index. |
| 104 ++next_insertion_index_; |
| 105 next_insertion_index_ %= kLookbackFrames; |
29 } | 106 } |
30 | 107 |
31 void ResidualEchoDetector::Initialize() { | 108 void ResidualEchoDetector::Initialize() { |
32 // TODO(ivoc): Add implementation. | 109 render_buffer_.Clear(); |
| 110 std::fill(render_power_.begin(), render_power_.end(), 0.f); |
| 111 std::fill(render_power_mean_.begin(), render_power_mean_.end(), 0.f); |
| 112 std::fill(render_power_std_dev_.begin(), render_power_std_dev_.end(), 0.f); |
| 113 render_statistics_.Clear(); |
| 114 capture_statistics_.Clear(); |
| 115 for (auto& cov : covariances_) { |
| 116 cov.Clear(); |
| 117 } |
| 118 echo_likelihood_ = 0.f; |
| 119 next_insertion_index_ = 0; |
33 } | 120 } |
34 | 121 |
35 void ResidualEchoDetector::PackRenderAudioBuffer( | 122 void ResidualEchoDetector::PackRenderAudioBuffer( |
36 AudioBuffer* audio, | 123 AudioBuffer* audio, |
37 std::vector<float>* packed_buffer) { | 124 std::vector<float>* packed_buffer) { |
38 RTC_DCHECK_GE(160u, audio->num_frames_per_band()); | 125 RTC_DCHECK_GE(160u, audio->num_frames_per_band()); |
39 | 126 |
40 packed_buffer->clear(); | 127 packed_buffer->clear(); |
41 packed_buffer->insert(packed_buffer->end(), | 128 packed_buffer->insert(packed_buffer->end(), |
42 audio->split_bands_const_f(0)[kBand0To8kHz], | 129 audio->split_bands_const_f(0)[kBand0To8kHz], |
43 (audio->split_bands_const_f(0)[kBand0To8kHz] + | 130 (audio->split_bands_const_f(0)[kBand0To8kHz] + |
44 audio->num_frames_per_band())); | 131 audio->num_frames_per_band())); |
45 } | 132 } |
46 | 133 |
47 float ResidualEchoDetector::get_echo_likelihood() const { | |
48 // TODO(ivoc): Add implementation. | |
49 return 0.0f; | |
50 } | |
51 | |
52 } // namespace webrtc | 134 } // namespace webrtc |
OLD | NEW |