Index: webrtc/modules/audio_processing/intelligibility/intelligibility_enhancer.h |
diff --git a/webrtc/modules/audio_processing/intelligibility/intelligibility_enhancer.h b/webrtc/modules/audio_processing/intelligibility/intelligibility_enhancer.h |
index 8125707f120981c40817152a80bcbea43f2e3006..d0818f688c5e6750b430dd81a429ccb3432a7f10 100644 |
--- a/webrtc/modules/audio_processing/intelligibility/intelligibility_enhancer.h |
+++ b/webrtc/modules/audio_processing/intelligibility/intelligibility_enhancer.h |
@@ -8,18 +8,14 @@ |
* be found in the AUTHORS file in the root of the source tree. |
*/ |
-// |
-// Specifies core class for intelligbility enhancement. |
-// |
- |
#ifndef WEBRTC_MODULES_AUDIO_PROCESSING_INTELLIGIBILITY_INTELLIGIBILITY_ENHANCER_H_ |
#define WEBRTC_MODULES_AUDIO_PROCESSING_INTELLIGIBILITY_INTELLIGIBILITY_ENHANCER_H_ |
#include <complex> |
-#include "webrtc/base/scoped_ptr.h" |
#include "webrtc/common_audio/lapped_transform.h" |
#include "webrtc/modules/audio_processing/intelligibility/intelligibility_utils.h" |
+#include "webrtc/system_wrappers/interface/scoped_ptr.h" |
struct WebRtcVadInst; |
typedef struct WebRtcVadInst VadInst; |
@@ -29,7 +25,6 @@ |
// Speech intelligibility enhancement module. Reads render and capture |
// audio streams and modifies the render stream with a set of gains per |
// frequency bin to enhance speech against the noise background. |
-// Note: assumes speech and noise streams are already separated. |
class IntelligibilityEnhancer { |
public: |
// Construct a new instance with the given filter bank resolution, |
@@ -38,43 +33,30 @@ |
// to elapse before a new gain computation is made. |variance_rate| specifies |
// the number of gain recomputations after which the variances are reset. |
// |cv_*| are parameters for the VarianceArray constructor for the |
- // clear speech stream. |
+ // lear speech stream. |
// TODO(bercic): the |cv_*|, |*_rate| and |gain_limit| parameters should |
// probably go away once fine tuning is done. They override the internal |
// constants in the class (kGainChangeLimit, kAnalyzeRate, kVarianceRate). |
- IntelligibilityEnhancer(int erb_resolution, |
- int sample_rate_hz, |
- int channels, |
- int cv_type, |
- float cv_alpha, |
- int cv_win, |
- int analysis_rate, |
- int variance_rate, |
+ IntelligibilityEnhancer(int erb_resolution, int sample_rate_hz, int channels, |
+ int cv_type, float cv_alpha, int cv_win, |
+ int analysis_rate, int variance_rate, |
float gain_limit); |
~IntelligibilityEnhancer(); |
- // Reads and processes chunk of noise stream in time domain. |
+ void ProcessRenderAudio(float* const* audio); |
void ProcessCaptureAudio(float* const* audio); |
- |
- // Reads chunk of speech in time domain and updates with modified signal. |
- void ProcessRenderAudio(float* const* audio); |
private: |
enum AudioSource { |
- kRenderStream = 0, // Clear speech stream. |
- kCaptureStream, // Noise stream. |
+ kRenderStream = 0, |
+ kCaptureStream, |
}; |
- // Provides access point to the frequency domain. |
class TransformCallback : public LappedTransform::Callback { |
public: |
TransformCallback(IntelligibilityEnhancer* parent, AudioSource source); |
- |
- // All in frequency domain, receives input |in_block|, applies |
- // intelligibility enhancement, and writes result to |out_block|. |
virtual void ProcessAudioBlock(const std::complex<float>* const* in_block, |
- int in_channels, |
- int frames, |
+ int in_channels, int frames, |
int out_channels, |
std::complex<float>* const* out_block); |
@@ -84,95 +66,72 @@ |
}; |
friend class TransformCallback; |
- // Sends streams to ProcessClearBlock or ProcessNoiseBlock based on source. |
- void DispatchAudio(AudioSource source, |
- const std::complex<float>* in_block, |
+ void DispatchAudio(AudioSource source, const std::complex<float>* in_block, |
std::complex<float>* out_block); |
- |
- // Updates variance computation and analysis with |in_block_|, |
- // and writes modified speech to |out_block|. |
void ProcessClearBlock(const std::complex<float>* in_block, |
std::complex<float>* out_block); |
- |
- // Computes and sets modified gains. |
void AnalyzeClearBlock(float power_target); |
- |
- // Updates variance calculation for noise input with |in_block|. |
void ProcessNoiseBlock(const std::complex<float>* in_block, |
std::complex<float>* out_block); |
- // Returns number of ERB filters. |
static int GetBankSize(int sample_rate, int erb_resolution); |
- |
- // Initializes ERB filterbank. |
void CreateErbBank(); |
- |
- // Analytically solves quadratic for optimal gains given |lambda|. |
- // Negative gains are set to 0. Stores the results in |sols|. |
- void SolveForGainsGivenLambda(float lambda, int start_freq, float* sols); |
- |
- // Computes variance across ERB filters from freq variance |var|. |
- // Stores in |result|. |
+ void SolveEquation14(float lambda, int start_freq, float* sols); |
void FilterVariance(const float* var, float* result); |
- |
- // Returns dot product of vectors specified by size |length| arrays |a|,|b|. |
static float DotProduct(const float* a, const float* b, int length); |
static const int kErbResolution; |
static const int kWindowSizeMs; |
static const int kChunkSizeMs; |
- static const int kAnalyzeRate; // Default for |analysis_rate_|. |
- static const int kVarianceRate; // Default for |variance_rate_|. |
+ static const int kAnalyzeRate; |
+ static const int kVarianceRate; |
static const float kClipFreq; |
- static const float kConfigRho; // Default production and interpretation SNR. |
+ static const float kConfigRho; |
static const float kKbdAlpha; |
static const float kGainChangeLimit; |
- const int freqs_; // Num frequencies in frequency domain. |
- const int window_size_; // Window size in samples; also the block size. |
- const int chunk_length_; // Chunk size in samples. |
- const int bank_size_; // Num ERB filters. |
+ const int freqs_; |
+ const int window_size_; // window size in samples; also the block size |
+ const int chunk_length_; // chunk size in samples |
+ const int bank_size_; |
const int sample_rate_hz_; |
const int erb_resolution_; |
- const int channels_; // Num channels. |
- const int analysis_rate_; // Num blocks before gains recalculated. |
- const int variance_rate_; // Num recalculations before history is cleared. |
+ const int channels_; |
+ const int analysis_rate_; |
+ const int variance_rate_; |
intelligibility::VarianceArray clear_variance_; |
intelligibility::VarianceArray noise_variance_; |
- rtc::scoped_ptr<float[]> filtered_clear_var_; |
- rtc::scoped_ptr<float[]> filtered_noise_var_; |
- float** filter_bank_; // TODO(ekmeyerson): Switch to using ChannelBuffer. |
- rtc::scoped_ptr<float[]> center_freqs_; |
+ scoped_ptr<float[]> filtered_clear_var_; |
+ scoped_ptr<float[]> filtered_noise_var_; |
+ float** filter_bank_; |
+ scoped_ptr<float[]> center_freqs_; |
int start_freq_; |
- rtc::scoped_ptr<float[]> rho_; // Production and interpretation SNR. |
- // for each ERB band. |
- rtc::scoped_ptr<float[]> gains_eq_; // Pre-filter modified gains. |
+ scoped_ptr<float[]> rho_; |
+ scoped_ptr<float[]> gains_eq_; |
intelligibility::GainApplier gain_applier_; |
// Destination buffer used to reassemble blocked chunks before overwriting |
// the original input array with modifications. |
- // TODO(ekmeyerson): Switch to using ChannelBuffer. |
float** temp_out_buffer_; |
- |
- rtc::scoped_ptr<float* []> input_audio_; |
- rtc::scoped_ptr<float[]> kbd_window_; |
+ scoped_ptr<float*[]> input_audio_; |
+ scoped_ptr<float[]> kbd_window_; |
TransformCallback render_callback_; |
TransformCallback capture_callback_; |
- rtc::scoped_ptr<LappedTransform> render_mangler_; |
- rtc::scoped_ptr<LappedTransform> capture_mangler_; |
+ scoped_ptr<LappedTransform> render_mangler_; |
+ scoped_ptr<LappedTransform> capture_mangler_; |
int block_count_; |
int analysis_step_; |
// TODO(bercic): Quick stopgap measure for voice detection in the clear |
// and noise streams. |
- // Note: VAD currently does not affect anything in IntelligibilityEnhancer. |
VadInst* vad_high_; |
VadInst* vad_low_; |
- rtc::scoped_ptr<int16_t[]> vad_tmp_buffer_; |
- bool has_voice_low_; // Whether voice detected in speech stream. |
+ scoped_ptr<int16_t[]> vad_tmp_buffer_; |
+ bool has_voice_low_; |
}; |
} // namespace webrtc |
#endif // WEBRTC_MODULES_AUDIO_PROCESSING_INTELLIGIBILITY_INTELLIGIBILITY_ENHANCER_H_ |
+ |