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 25 matching lines...) Expand all Loading... | |
36 // Construct a new instance with the given filter bank resolution, | 36 // Construct a new instance with the given filter bank resolution, |
37 // sampling rate, number of channels and analysis rates. | 37 // sampling rate, number of channels and analysis rates. |
38 // |analysis_rate| sets the number of input blocks (containing speech!) | 38 // |analysis_rate| sets the number of input blocks (containing speech!) |
39 // to elapse before a new gain computation is made. |variance_rate| specifies | 39 // to elapse before a new gain computation is made. |variance_rate| specifies |
40 // the number of gain recomputations after which the variances are reset. | 40 // the number of gain recomputations after which the variances are reset. |
41 // |cv_*| are parameters for the VarianceArray constructor for the | 41 // |cv_*| are parameters for the VarianceArray constructor for the |
42 // clear speech stream. | 42 // clear speech stream. |
43 // TODO(bercic): the |cv_*|, |*_rate| and |gain_limit| parameters should | 43 // TODO(bercic): the |cv_*|, |*_rate| and |gain_limit| parameters should |
44 // probably go away once fine tuning is done. They override the internal | 44 // probably go away once fine tuning is done. They override the internal |
45 // constants in the class (kGainChangeLimit, kAnalyzeRate, kVarianceRate). | 45 // constants in the class (kGainChangeLimit, kAnalyzeRate, kVarianceRate). |
46 IntelligibilityEnhancer(int erb_resolution, | 46 IntelligibilityEnhancer(size_t erb_resolution, |
47 int sample_rate_hz, | 47 int sample_rate_hz, |
48 int channels, | 48 int channels, |
Andrew MacDonald
2015/07/24 04:01:44
size_t
| |
49 int cv_type, | 49 int cv_type, |
50 float cv_alpha, | 50 float cv_alpha, |
51 int cv_win, | 51 size_t cv_win, |
52 int analysis_rate, | 52 int analysis_rate, |
53 int variance_rate, | 53 int variance_rate, |
54 float gain_limit); | 54 float gain_limit); |
55 ~IntelligibilityEnhancer(); | 55 ~IntelligibilityEnhancer(); |
56 | 56 |
57 // Reads and processes chunk of noise stream in time domain. | 57 // Reads and processes chunk of noise stream in time domain. |
58 void ProcessCaptureAudio(float* const* audio); | 58 void ProcessCaptureAudio(float* const* audio); |
59 | 59 |
60 // Reads chunk of speech in time domain and updates with modified signal. | 60 // Reads chunk of speech in time domain and updates with modified signal. |
61 void ProcessRenderAudio(float* const* audio); | 61 void ProcessRenderAudio(float* const* audio); |
62 | 62 |
63 private: | 63 private: |
64 enum AudioSource { | 64 enum AudioSource { |
65 kRenderStream = 0, // Clear speech stream. | 65 kRenderStream = 0, // Clear speech stream. |
66 kCaptureStream, // Noise stream. | 66 kCaptureStream, // Noise stream. |
67 }; | 67 }; |
68 | 68 |
69 // Provides access point to the frequency domain. | 69 // Provides access point to the frequency domain. |
70 class TransformCallback : public LappedTransform::Callback { | 70 class TransformCallback : public LappedTransform::Callback { |
71 public: | 71 public: |
72 TransformCallback(IntelligibilityEnhancer* parent, AudioSource source); | 72 TransformCallback(IntelligibilityEnhancer* parent, AudioSource source); |
73 | 73 |
74 // All in frequency domain, receives input |in_block|, applies | 74 // All in frequency domain, receives input |in_block|, applies |
75 // intelligibility enhancement, and writes result to |out_block|. | 75 // intelligibility enhancement, and writes result to |out_block|. |
76 void ProcessAudioBlock(const std::complex<float>* const* in_block, | 76 void ProcessAudioBlock(const std::complex<float>* const* in_block, |
77 int in_channels, | 77 int in_channels, |
78 int frames, | 78 size_t frames, |
79 int out_channels, | 79 int out_channels, |
80 std::complex<float>* const* out_block) override; | 80 std::complex<float>* const* out_block) override; |
81 | 81 |
82 private: | 82 private: |
83 IntelligibilityEnhancer* parent_; | 83 IntelligibilityEnhancer* parent_; |
84 AudioSource source_; | 84 AudioSource source_; |
85 }; | 85 }; |
86 friend class TransformCallback; | 86 friend class TransformCallback; |
87 FRIEND_TEST_ALL_PREFIXES(IntelligibilityEnhancerTest, TestErbCreation); | 87 FRIEND_TEST_ALL_PREFIXES(IntelligibilityEnhancerTest, TestErbCreation); |
88 FRIEND_TEST_ALL_PREFIXES(IntelligibilityEnhancerTest, TestSolveForGains); | 88 FRIEND_TEST_ALL_PREFIXES(IntelligibilityEnhancerTest, TestSolveForGains); |
(...skipping 15 matching lines...) Expand all Loading... | |
104 void SolveForLambda(float power_target, float power_bot, float power_top); | 104 void SolveForLambda(float power_target, float power_bot, float power_top); |
105 | 105 |
106 // Transforms freq gains to ERB gains. | 106 // Transforms freq gains to ERB gains. |
107 void UpdateErbGains(); | 107 void UpdateErbGains(); |
108 | 108 |
109 // Updates variance calculation for noise input with |in_block|. | 109 // Updates variance calculation for noise input with |in_block|. |
110 void ProcessNoiseBlock(const std::complex<float>* in_block, | 110 void ProcessNoiseBlock(const std::complex<float>* in_block, |
111 std::complex<float>* out_block); | 111 std::complex<float>* out_block); |
112 | 112 |
113 // Returns number of ERB filters. | 113 // Returns number of ERB filters. |
114 static int GetBankSize(int sample_rate, int erb_resolution); | 114 static size_t GetBankSize(int sample_rate, size_t erb_resolution); |
115 | 115 |
116 // Initializes ERB filterbank. | 116 // Initializes ERB filterbank. |
117 void CreateErbBank(); | 117 void CreateErbBank(); |
118 | 118 |
119 // Analytically solves quadratic for optimal gains given |lambda|. | 119 // Analytically solves quadratic for optimal gains given |lambda|. |
120 // Negative gains are set to 0. Stores the results in |sols|. | 120 // Negative gains are set to 0. Stores the results in |sols|. |
121 void SolveForGainsGivenLambda(float lambda, int start_freq, float* sols); | 121 void SolveForGainsGivenLambda(float lambda, size_t start_freq, float* sols); |
122 | 122 |
123 // Computes variance across ERB filters from freq variance |var|. | 123 // Computes variance across ERB filters from freq variance |var|. |
124 // Stores in |result|. | 124 // Stores in |result|. |
125 void FilterVariance(const float* var, float* result); | 125 void FilterVariance(const float* var, float* result); |
126 | 126 |
127 // Returns dot product of vectors specified by size |length| arrays |a|,|b|. | 127 // Returns dot product of vectors specified by size |length| arrays |a|,|b|. |
128 static float DotProduct(const float* a, const float* b, int length); | 128 static float DotProduct(const float* a, const float* b, size_t length); |
129 | 129 |
130 const int freqs_; // Num frequencies in frequency domain. | 130 const size_t freqs_; // Num frequencies in frequency domain. |
131 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. |
132 const int chunk_length_; // Chunk size in samples. | 132 const size_t chunk_length_; // Chunk size in samples. |
133 const int bank_size_; // Num ERB filters. | 133 const size_t bank_size_; // Num ERB filters. |
134 const int sample_rate_hz_; | 134 const int sample_rate_hz_; |
135 const int erb_resolution_; | 135 const int erb_resolution_; |
136 const int channels_; // Num channels. | 136 const int channels_; // Num channels. |
137 const int analysis_rate_; // Num blocks before gains recalculated. | 137 const int analysis_rate_; // Num blocks before gains recalculated. |
138 const int variance_rate_; // Num recalculations before history is cleared. | 138 const int variance_rate_; // Num recalculations before history is cleared. |
139 | 139 |
140 intelligibility::VarianceArray clear_variance_; | 140 intelligibility::VarianceArray clear_variance_; |
141 intelligibility::VarianceArray noise_variance_; | 141 intelligibility::VarianceArray noise_variance_; |
142 rtc::scoped_ptr<float[]> filtered_clear_var_; | 142 rtc::scoped_ptr<float[]> filtered_clear_var_; |
143 rtc::scoped_ptr<float[]> filtered_noise_var_; | 143 rtc::scoped_ptr<float[]> filtered_noise_var_; |
144 std::vector<std::vector<float>> filter_bank_; | 144 std::vector<std::vector<float>> filter_bank_; |
145 rtc::scoped_ptr<float[]> center_freqs_; | 145 rtc::scoped_ptr<float[]> center_freqs_; |
146 int start_freq_; | 146 size_t start_freq_; |
147 rtc::scoped_ptr<float[]> rho_; // Production and interpretation SNR. | 147 rtc::scoped_ptr<float[]> rho_; // Production and interpretation SNR. |
148 // for each ERB band. | 148 // for each ERB band. |
149 rtc::scoped_ptr<float[]> gains_eq_; // Pre-filter modified gains. | 149 rtc::scoped_ptr<float[]> gains_eq_; // Pre-filter modified gains. |
150 intelligibility::GainApplier gain_applier_; | 150 intelligibility::GainApplier gain_applier_; |
151 | 151 |
152 // Destination buffer used to reassemble blocked chunks before overwriting | 152 // Destination buffer used to reassemble blocked chunks before overwriting |
153 // the original input array with modifications. | 153 // the original input array with modifications. |
154 // TODO(ekmeyerson): Switch to using ChannelBuffer. | 154 // TODO(ekmeyerson): Switch to using ChannelBuffer. |
155 float** temp_out_buffer_; | 155 float** temp_out_buffer_; |
156 | 156 |
(...skipping 11 matching lines...) Expand all Loading... | |
168 // Note: VAD currently does not affect anything in IntelligibilityEnhancer. | 168 // Note: VAD currently does not affect anything in IntelligibilityEnhancer. |
169 VadInst* vad_high_; | 169 VadInst* vad_high_; |
170 VadInst* vad_low_; | 170 VadInst* vad_low_; |
171 rtc::scoped_ptr<int16_t[]> vad_tmp_buffer_; | 171 rtc::scoped_ptr<int16_t[]> vad_tmp_buffer_; |
172 bool has_voice_low_; // Whether voice detected in speech stream. | 172 bool has_voice_low_; // Whether voice detected in speech stream. |
173 }; | 173 }; |
174 | 174 |
175 } // namespace webrtc | 175 } // namespace webrtc |
176 | 176 |
177 #endif // WEBRTC_MODULES_AUDIO_PROCESSING_INTELLIGIBILITY_INTELLIGIBILITY_ENHAN CER_H_ | 177 #endif // WEBRTC_MODULES_AUDIO_PROCESSING_INTELLIGIBILITY_INTELLIGIBILITY_ENHAN CER_H_ |
OLD | NEW |