webrtc回音消除模块

/* * Copyright (c) 2012 The WebRTC project authors. All Rights Reserved. * * Use of this source code is governed by a BSD-style license * that can be found in the LICENSE file in the root of the source * tree. An additional intellectual property rights grant can be found * in the file PATENTS. All contributing project authors may * be found in the AUTHORS file in the root of the source tree. */ #include <math.h> #include <string.h> #include <stdlib.h> #include "rtc_base/checks.h" #include "common_audio/signal_processing/include/signal_processing_library.h" #include "common_audio/third_party/fft4g/fft4g.h" #include "modules/audio_processing/ns/noise_suppression.h" #include "modules/audio_processing/ns/ns_core.h" #include "modules/audio_processing/ns/windows_private.h" // Set Feature Extraction Parameters. static void set_feature_extraction_parameters(NoiseSuppressionC* self) { // Bin size of histogram. self->featureExtractionParams.binSizeLrt = 0.1f; self->featureExtractionParams.binSizeSpecFlat = 0.05f; self->featureExtractionParams.binSizeSpecDiff = 0.1f; // Range of histogram over which LRT threshold is computed. self->featureExtractionParams.rangeAvgHistLrt = 1.f; // Scale parameters: multiply dominant peaks of the histograms by scale factor // to obtain thresholds for prior model. // For LRT and spectral difference. self->featureExtractionParams.factor1ModelPars = 1.2f; // For spectral_flatness: used when noise is flatter than speech. self->featureExtractionParams.factor2ModelPars = 0.9f; // Peak limit for spectral flatness (varies between 0 and 1). self->featureExtractionParams.thresPosSpecFlat = 0.6f; // Limit on spacing of two highest peaks in histogram: spacing determined by // bin size. self->featureExtractionParams.limitPeakSpacingSpecFlat = 2 * self->featureExtractionParams.binSizeSpecFlat; self->featureExtractionParams.limitPeakSpacingSpecDiff = 2 * self->featureExtractionParams.binSizeSpecDiff; // Limit on relevance of second peak. self->featureExtractionParams.limitPeakWeightsSpecFlat = 0.5f; self->featureExtractionParams.limitPeakWeightsSpecDiff = 0.5f; // Fluctuation limit of LRT feature. self->featureExtractionParams.thresFluctLrt = 0.05f; // Limit on the max and min values for the feature thresholds. self->featureExtractionParams.maxLrt = 1.f; self->featureExtractionParams.minLrt = 0.2f; self->featureExtractionParams.maxSpecFlat = 0.95f; self->featureExtractionParams.minSpecFlat = 0.1f; self->featureExtractionParams.maxSpecDiff = 1.f; self->featureExtractionParams.minSpecDiff = 0.16f; // Criteria of weight of histogram peak to accept/reject feature. self->featureExtractionParams.thresWeightSpecFlat = (int)(0.3 * (self->modelUpdatePars[1])); // For spectral flatness. self->featureExtractionParams.thresWeightSpecDiff = (int)(0.3 * (self->modelUpdatePars[1])); // For spectral difference. } // Initialize state. int WebRtcNs_InitCore(NoiseSuppressionC* self, uint32_t fs) { int i; // Check for valid pointer. if (self == NULL) { return -1; } // Initialization of struct. if (fs == 8000 || fs == 16000 || fs == 32000 || fs == 48000) { self->fs = fs; } else { return -1; } self->windShift = 0; // We only support 10ms frames. if (fs == 8000) { self->blockLen = 80; self->anaLen = 128; self->window = kBlocks80w128; } else { self->blockLen = 160; self->anaLen = 256; self->window = kBlocks160w256; } self->magnLen = self->anaLen / 2 + 1; // Number of frequency bins. // Initialize FFT work arrays. self->ip[0] = 0; // Setting this triggers initialization. memset(self->dataBuf, 0, sizeof(float) * ANAL_BLOCKL_MAX); WebRtc_rdft(self->anaLen, 1, self->dataBuf, self->ip, self->wfft); memset(self->analyzeBuf, 0, sizeof(float) * ANAL_BLOCKL_MAX); memset(self->dataBuf, 0, sizeof(float) * ANAL_BLOCKL_MAX); memset(self->syntBuf, 0, sizeof(float) * ANAL_BLOCKL_MAX); // For HB processing. memset(self->dataBufHB, 0, sizeof(float) * NUM_HIGH_BANDS_MAX * ANAL_BLOCKL_MAX); // For quantile noise estimation. memset(self->quantile, 0, sizeof(float) * HALF_ANAL_BLOCKL); for (i = 0; i < SIMULT * HALF_ANAL_BLOCKL; i++) { self->lquantile[i] = 8.f; self->density[i] = 0.3f; } for (i = 0; i < SIMULT; i++) { self->counter[i] = (int)floor((float)(END_STARTUP_LONG * (i + 1)) / (float)SIMULT); } self->updates = 0; // Wiener filter initialization. for (i = 0; i < HALF_ANAL_BLOCKL; i++) { self->smooth[i] = 1.f; } // Set the aggressiveness: default. self->aggrMode = 0; // Initialize variables for new method. self->priorSpeechProb = 0.5f; // Prior prob for speech/noise. // Previous analyze mag spectrum. memset(self->magnPrevAnalyze, 0, sizeof(float) * HALF_ANAL_BLOCKL); // Previous process mag spectrum. memset(self->magnPrevProcess, 0, sizeof(float) * HALF_ANAL_BLOCKL); // Current noise-spectrum. memset(self->noise, 0, sizeof(float) * HALF_ANAL_BLOCKL); // Previous noise-spectrum. memset(self->noisePrev, 0, sizeof(float) * HALF_ANAL_BLOCKL); // Conservative noise spectrum estimate. memset(self->magnAvgPause, 0, sizeof(float) * HALF_ANAL_BLOCKL); // For estimation of HB in second pass. memset(self->speechProb, 0, sizeof(float) * HALF_ANAL_BLOCKL); // Initial average magnitude spectrum. memset(self->initMagnEst, 0, sizeof(float) * HALF_ANAL_BLOCKL); for (i = 0; i < HALF_ANAL_BLOCKL; i++) { // Smooth LR (same as threshold). self->logLrtTimeAvg[i] = LRT_FEATURE_THR; } // Feature quantities. // Spectral flatness (start on threshold). self->featureData[0] = SF_FEATURE_THR; self->featureData[1] = 0.f; // Spectral entropy: not used in this version. self->featureData[2] = 0.f; // Spectral variance: not used in this version. // Average LRT factor (start on threshold). self->featureData[3] = LRT_FEATURE_THR; // Spectral template diff (start on threshold). self->featureData[4] = SF_FEATURE_THR; self->featureData[5] = 0.f; // Normalization for spectral difference. // Window time-average of input magnitude spectrum. self->featureData[6] = 0.f; memset(self->parametricNoise, 0, sizeof(float) * HALF_ANAL_BLOCKL); // Histogram quantities: used to estimate/update thresholds for features. memset(self->histLrt, 0, sizeof(int) * HIST_PAR_EST); memset(self->histSpecFlat, 0, sizeof(int) * HIST_PAR_EST); memset(self->histSpecDiff, 0, sizeof(int) * HIST_PAR_EST); self->blockInd = -1; // Frame counter. // Default threshold for LRT feature. self->priorModelPars[0] = LRT_FEATURE_THR; // Threshold for spectral flatness: determined on-line. self->priorModelPars[1] = 0.5f; // sgn_map par for spectral measure: 1 for flatness measure. self->priorModelPars[2] = 1.f; // Threshold for template-difference feature: determined on-line. self->priorModelPars[3] = 0.5f; // Default weighting parameter for LRT feature. self->priorModelPars[4] = 1.f; // Default weighting parameter for spectral flatness feature. self->priorModelPars[5] = 0.f; // Default weighting parameter for spectral difference feature. self->priorModelPars[6] = 0.f; // Update flag for parameters: // 0 no update, 1 = update once, 2 = update every window. self->modelUpdatePars[0] = 2; self->modelUpdatePars[1] = 500; // Window for update. // Counter for update of conservative noise spectrum. self->modelUpdatePars[2] = 0; // Counter if the feature thresholds are updated during the sequence. self->modelUpdatePars[3] = self->modelUpdatePars[1]; self->signalEnergy = 0.0; self->sumMagn = 0.0; self->whiteNoiseLevel = 0.0; self->pinkNoiseNumerator = 0.0; self->pinkNoiseExp = 0.0; set_feature_extraction_parameters(self); // Default mode. WebRtcNs_set_policy_core(self, 0); self->initFlag = 1; return 0; } // Estimate noise. static void Nois