OLD | NEW |
1 /* | 1 /* |
2 * Copyright (c) 2014 The WebRTC project authors. All Rights Reserved. | 2 * Copyright (c) 2014 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 |
(...skipping 24 matching lines...) Expand all Loading... |
35 // Construct a new instance with the given filter bank resolution, | 35 // Construct a new instance with the given filter bank resolution, |
36 // sampling rate, number of channels and analysis rates. | 36 // sampling rate, number of channels and analysis rates. |
37 // |analysis_rate| sets the number of input blocks (containing speech!) | 37 // |analysis_rate| sets the number of input blocks (containing speech!) |
38 // to elapse before a new gain computation is made. |variance_rate| specifies | 38 // to elapse before a new gain computation is made. |variance_rate| specifies |
39 // the number of gain recomputations after which the variances are reset. | 39 // the number of gain recomputations after which the variances are reset. |
40 // |cv_*| are parameters for the VarianceArray constructor for the | 40 // |cv_*| are parameters for the VarianceArray constructor for the |
41 // clear speech stream. | 41 // clear speech stream. |
42 // TODO(bercic): the |cv_*|, |*_rate| and |gain_limit| parameters should | 42 // TODO(bercic): the |cv_*|, |*_rate| and |gain_limit| parameters should |
43 // probably go away once fine tuning is done. They override the internal | 43 // probably go away once fine tuning is done. They override the internal |
44 // constants in the class (kGainChangeLimit, kAnalyzeRate, kVarianceRate). | 44 // constants in the class (kGainChangeLimit, kAnalyzeRate, kVarianceRate). |
45 IntelligibilityEnhancer(int erb_resolution, | 45 IntelligibilityEnhancer(size_t erb_resolution, |
46 int sample_rate_hz, | 46 int sample_rate_hz, |
47 int channels, | 47 int channels, |
48 int cv_type, | 48 int cv_type, |
49 float cv_alpha, | 49 float cv_alpha, |
50 int cv_win, | 50 size_t cv_win, |
51 int analysis_rate, | 51 int analysis_rate, |
52 int variance_rate, | 52 int variance_rate, |
53 float gain_limit); | 53 float gain_limit); |
54 ~IntelligibilityEnhancer(); | 54 ~IntelligibilityEnhancer(); |
55 | 55 |
56 // Reads and processes chunk of noise stream in time domain. | 56 // Reads and processes chunk of noise stream in time domain. |
57 void ProcessCaptureAudio(float* const* audio); | 57 void ProcessCaptureAudio(float* const* audio); |
58 | 58 |
59 // Reads chunk of speech in time domain and updates with modified signal. | 59 // Reads chunk of speech in time domain and updates with modified signal. |
60 void ProcessRenderAudio(float* const* audio); | 60 void ProcessRenderAudio(float* const* audio); |
61 | 61 |
62 private: | 62 private: |
63 enum AudioSource { | 63 enum AudioSource { |
64 kRenderStream = 0, // Clear speech stream. | 64 kRenderStream = 0, // Clear speech stream. |
65 kCaptureStream, // Noise stream. | 65 kCaptureStream, // Noise stream. |
66 }; | 66 }; |
67 | 67 |
68 // Provides access point to the frequency domain. | 68 // Provides access point to the frequency domain. |
69 class TransformCallback : public LappedTransform::Callback { | 69 class TransformCallback : public LappedTransform::Callback { |
70 public: | 70 public: |
71 TransformCallback(IntelligibilityEnhancer* parent, AudioSource source); | 71 TransformCallback(IntelligibilityEnhancer* parent, AudioSource source); |
72 | 72 |
73 // All in frequency domain, receives input |in_block|, applies | 73 // All in frequency domain, receives input |in_block|, applies |
74 // intelligibility enhancement, and writes result to |out_block|. | 74 // intelligibility enhancement, and writes result to |out_block|. |
75 virtual void ProcessAudioBlock(const std::complex<float>* const* in_block, | 75 void ProcessAudioBlock(const std::complex<float>* const* in_block, |
76 int in_channels, | 76 int in_channels, |
77 int frames, | 77 size_t frames, |
78 int out_channels, | 78 int out_channels, |
79 std::complex<float>* const* out_block); | 79 std::complex<float>* const* out_block) override; |
80 | 80 |
81 private: | 81 private: |
82 IntelligibilityEnhancer* parent_; | 82 IntelligibilityEnhancer* parent_; |
83 AudioSource source_; | 83 AudioSource source_; |
84 }; | 84 }; |
85 friend class TransformCallback; | 85 friend class TransformCallback; |
86 | 86 |
87 // Sends streams to ProcessClearBlock or ProcessNoiseBlock based on source. | 87 // Sends streams to ProcessClearBlock or ProcessNoiseBlock based on source. |
88 void DispatchAudio(AudioSource source, | 88 void DispatchAudio(AudioSource source, |
89 const std::complex<float>* in_block, | 89 const std::complex<float>* in_block, |
90 std::complex<float>* out_block); | 90 std::complex<float>* out_block); |
91 | 91 |
92 // Updates variance computation and analysis with |in_block_|, | 92 // Updates variance computation and analysis with |in_block_|, |
93 // and writes modified speech to |out_block|. | 93 // and writes modified speech to |out_block|. |
94 void ProcessClearBlock(const std::complex<float>* in_block, | 94 void ProcessClearBlock(const std::complex<float>* in_block, |
95 std::complex<float>* out_block); | 95 std::complex<float>* out_block); |
96 | 96 |
97 // Computes and sets modified gains. | 97 // Computes and sets modified gains. |
98 void AnalyzeClearBlock(float power_target); | 98 void AnalyzeClearBlock(float power_target); |
99 | 99 |
100 // Updates variance calculation for noise input with |in_block|. | 100 // Updates variance calculation for noise input with |in_block|. |
101 void ProcessNoiseBlock(const std::complex<float>* in_block, | 101 void ProcessNoiseBlock(const std::complex<float>* in_block, |
102 std::complex<float>* out_block); | 102 std::complex<float>* out_block); |
103 | 103 |
104 // Returns number of ERB filters. | 104 // Returns number of ERB filters. |
105 static int GetBankSize(int sample_rate, int erb_resolution); | 105 static size_t GetBankSize(int sample_rate, size_t erb_resolution); |
106 | 106 |
107 // Initializes ERB filterbank. | 107 // Initializes ERB filterbank. |
108 void CreateErbBank(); | 108 void CreateErbBank(); |
109 | 109 |
110 // Analytically solves quadratic for optimal gains given |lambda|. | 110 // Analytically solves quadratic for optimal gains given |lambda|. |
111 // Negative gains are set to 0. Stores the results in |sols|. | 111 // Negative gains are set to 0. Stores the results in |sols|. |
112 void SolveForGainsGivenLambda(float lambda, int start_freq, float* sols); | 112 void SolveForGainsGivenLambda(float lambda, size_t start_freq, float* sols); |
113 | 113 |
114 // Computes variance across ERB filters from freq variance |var|. | 114 // Computes variance across ERB filters from freq variance |var|. |
115 // Stores in |result|. | 115 // Stores in |result|. |
116 void FilterVariance(const float* var, float* result); | 116 void FilterVariance(const float* var, float* result); |
117 | 117 |
118 // Returns dot product of vectors specified by size |length| arrays |a|,|b|. | 118 // Returns dot product of vectors specified by size |length| arrays |a|,|b|. |
119 static float DotProduct(const float* a, const float* b, int length); | 119 static float DotProduct(const float* a, const float* b, size_t length); |
120 | 120 |
121 static const int kErbResolution; | |
122 static const int kWindowSizeMs; | 121 static const int kWindowSizeMs; |
123 static const int kChunkSizeMs; | 122 static const int kChunkSizeMs; |
124 static const int kAnalyzeRate; // Default for |analysis_rate_|. | 123 static const int kAnalyzeRate; // Default for |analysis_rate_|. |
125 static const int kVarianceRate; // Default for |variance_rate_|. | 124 static const int kVarianceRate; // Default for |variance_rate_|. |
126 static const float kClipFreq; | 125 static const float kClipFreq; |
127 static const float kConfigRho; // Default production and interpretation SNR. | 126 static const float kConfigRho; // Default production and interpretation SNR. |
128 static const float kKbdAlpha; | 127 static const float kKbdAlpha; |
129 static const float kGainChangeLimit; | 128 static const float kGainChangeLimit; |
130 | 129 |
131 const int freqs_; // Num frequencies in frequency domain. | 130 const size_t freqs_; // Num frequencies in frequency domain. |
132 const int window_size_; // Window size in samples; also the block size. | 131 const size_t window_size_; // Window size in samples; also the block size. |
133 const int chunk_length_; // Chunk size in samples. | 132 const size_t chunk_length_; // Chunk size in samples. |
134 const int bank_size_; // Num ERB filters. | 133 const size_t bank_size_; // Num ERB filters. |
135 const int sample_rate_hz_; | 134 const int sample_rate_hz_; |
136 const int erb_resolution_; | 135 const int erb_resolution_; |
137 const int channels_; // Num channels. | 136 const int channels_; // Num channels. |
138 const int analysis_rate_; // Num blocks before gains recalculated. | 137 const int analysis_rate_; // Num blocks before gains recalculated. |
139 const int variance_rate_; // Num recalculations before history is cleared. | 138 const int variance_rate_; // Num recalculations before history is cleared. |
140 | 139 |
141 intelligibility::VarianceArray clear_variance_; | 140 intelligibility::VarianceArray clear_variance_; |
142 intelligibility::VarianceArray noise_variance_; | 141 intelligibility::VarianceArray noise_variance_; |
143 rtc::scoped_ptr<float[]> filtered_clear_var_; | 142 rtc::scoped_ptr<float[]> filtered_clear_var_; |
144 rtc::scoped_ptr<float[]> filtered_noise_var_; | 143 rtc::scoped_ptr<float[]> filtered_noise_var_; |
145 float** filter_bank_; // TODO(ekmeyerson): Switch to using ChannelBuffer. | 144 float** filter_bank_; // TODO(ekmeyerson): Switch to using ChannelBuffer. |
146 rtc::scoped_ptr<float[]> center_freqs_; | 145 rtc::scoped_ptr<float[]> center_freqs_; |
147 int start_freq_; | 146 size_t start_freq_; |
148 rtc::scoped_ptr<float[]> rho_; // Production and interpretation SNR. | 147 rtc::scoped_ptr<float[]> rho_; // Production and interpretation SNR. |
149 // for each ERB band. | 148 // for each ERB band. |
150 rtc::scoped_ptr<float[]> gains_eq_; // Pre-filter modified gains. | 149 rtc::scoped_ptr<float[]> gains_eq_; // Pre-filter modified gains. |
151 intelligibility::GainApplier gain_applier_; | 150 intelligibility::GainApplier gain_applier_; |
152 | 151 |
153 // Destination buffer used to reassemble blocked chunks before overwriting | 152 // Destination buffer used to reassemble blocked chunks before overwriting |
154 // the original input array with modifications. | 153 // the original input array with modifications. |
155 // TODO(ekmeyerson): Switch to using ChannelBuffer. | 154 // TODO(ekmeyerson): Switch to using ChannelBuffer. |
156 float** temp_out_buffer_; | 155 float** temp_out_buffer_; |
157 | 156 |
(...skipping 11 matching lines...) Expand all Loading... |
169 // Note: VAD currently does not affect anything in IntelligibilityEnhancer. | 168 // Note: VAD currently does not affect anything in IntelligibilityEnhancer. |
170 VadInst* vad_high_; | 169 VadInst* vad_high_; |
171 VadInst* vad_low_; | 170 VadInst* vad_low_; |
172 rtc::scoped_ptr<int16_t[]> vad_tmp_buffer_; | 171 rtc::scoped_ptr<int16_t[]> vad_tmp_buffer_; |
173 bool has_voice_low_; // Whether voice detected in speech stream. | 172 bool has_voice_low_; // Whether voice detected in speech stream. |
174 }; | 173 }; |
175 | 174 |
176 } // namespace webrtc | 175 } // namespace webrtc |
177 | 176 |
178 #endif // WEBRTC_MODULES_AUDIO_PROCESSING_INTELLIGIBILITY_INTELLIGIBILITY_ENHAN
CER_H_ | 177 #endif // WEBRTC_MODULES_AUDIO_PROCESSING_INTELLIGIBILITY_INTELLIGIBILITY_ENHAN
CER_H_ |
OLD | NEW |