音頻自動增益 與 靜音檢測 演算法 附完整C代碼【轉】

標籤：tar   output   cpu   rate   放大   als   .com   類比   溢出   

轉自：https://www.cnblogs.com/cpuimage/p/8908551.html

前面分享過一個演算法《音頻增益響度分析 ReplayGain 附完整C程式碼範例》

主要用於評估一定長度音訊音量強度，

而分析之後，很多類似的需求，肯定是做音頻增益，提高音量諸如此類做法。

不過在項目實測的時候，其實真的很難定標準，

到底在什麼樣的環境下，要增大音量，還是降低。

在通訊行業一般的做法就是採用靜音檢測，

一旦檢測為靜音或者噪音，則不做處理，反之通過一定的策略進行處理。

這裡就涉及到兩個演算法，一個是靜音檢測，一個是音頻增益。

增益其實沒什麼好說的，類似於資料歸一化展開的做法。

靜音檢測 在WebRTC中 是採用計算GMM (Gaussian Mixture Model,高斯混合模型)進行特徵提取的。

在很長一段時間裡面，音頻特徵 有3個主要的方法，

GMM  ,Spectrogram (聲譜圖), MFCC 即 Mel-Frequency Cepstrum(Mel頻率倒譜)

恕我直言,GMM 提取的特徵，其魯棒性 不如後兩者。

也不多做介紹，感興趣的同學，翻翻 維基百科 ,補補課。

當然在實際使用演算法時，會由此延伸出來一些小技巧。

例如，用靜音檢測 來做音頻裁剪，或者搭配音頻增益做一些音頻增強之類的操作。

自動增益在WebRTC 原始碼檔案是:analog_agc.c 和 digital_agc.c

靜音檢測 原始碼檔案是: webrtc_vad.c

這個命名，有一定的曆史原因了。

經過梳理後，

增益演算法為 agc.c agc.h

靜音檢測為 vad.c vad.h

增益演算法的完整範例程式碼：

#include <stdio.h>#include <stdlib.h>#include <stdint.h>//採用https://github.com/mackron/dr_libs/blob/master/dr_wav.h 解碼#define DR_WAV_IMPLEMENTATION#include "dr_wav.h"#include "agc.h"#ifndef nullptr#define nullptr 0#endif#ifndef MIN#define  MIN(A, B)        ((A) < (B) ? (A) : (B))#endif//寫wav檔案void wavWrite_int16(char *filename, int16_t *buffer, size_t sampleRate, size_t totalSampleCount) {    drwav_data_format format = {};    format.container = drwav_container_riff;     // <-- drwav_container_riff = normal WAV files, drwav_container_w64 = Sony Wave64.    format.format = DR_WAVE_FORMAT_PCM;          // <-- Any of the DR_WAVE_FORMAT_* codes.    format.channels = 1;    format.sampleRate = (drwav_uint32) sampleRate;    format.bitsPerSample = 16;    drwav *pWav = drwav_open_file_write(filename, &format);    if (pWav) {        drwav_uint64 samplesWritten = drwav_write(pWav, totalSampleCount, buffer);        drwav_uninit(pWav);        if (samplesWritten != totalSampleCount) {            fprintf(stderr, "ERROR\n");            exit(1);        }    }}//讀取wav檔案int16_t *wavRead_int16(char *filename, uint32_t *sampleRate, uint64_t *totalSampleCount) {    unsigned int channels;    int16_t *buffer = drwav_open_and_read_file_s16(filename, &channels, sampleRate, totalSampleCount);    if (buffer == nullptr) {        printf("讀取wav檔案失敗.");    }    //僅僅處理單通道音頻    if (channels != 1) {        drwav_free(buffer);        buffer = nullptr;        *sampleRate = 0;        *totalSampleCount = 0;    }    return buffer;}//分割路徑函數void splitpath(const char *path, char *drv, char *dir, char *name, char *ext) {    const char *end;    const char *p;    const char *s;    if (path[0] && path[1] == ‘:‘) {        if (drv) {            *drv++ = *path++;            *drv++ = *path++;            *drv = ‘\0‘;        }    } else if (drv)        *drv = ‘\0‘;    for (end = path; *end && *end != ‘:‘;)        end++;    for (p = end; p > path && *--p != ‘\\‘ && *p != ‘/‘;)        if (*p == ‘.‘) {            end = p;            break;        }    if (ext)        for (s = end; (*ext = *s++);)            ext++;    for (p = end; p > path;)        if (*--p == ‘\\‘ || *p == ‘/‘) {            p++;            break;        }    if (name) {        for (s = p; s < end;)            *name++ = *s++;        *name = ‘\0‘;    }    if (dir) {        for (s = path; s < p;)            *dir++ = *s++;        *dir = ‘\0‘;    }}int agcProcess(int16_t *buffer, uint32_t sampleRate, size_t samplesCount, int16_t agcMode) {    if (buffer == nullptr) return -1;    if (samplesCount == 0) return -1;    WebRtcAgcConfig agcConfig;    agcConfig.compressionGaindB = 9; // default 9 dB    agcConfig.limiterEnable = 1; // default kAgcTrue (on)    agcConfig.targetLevelDbfs = 3; // default 3 (-3 dBOv)    int minLevel = 0;    int maxLevel = 255;    size_t samples = MIN(160, sampleRate / 100);    if (samples == 0) return -1;    const int maxSamples = 320;    int16_t *input = buffer;    size_t nTotal = (samplesCount / samples);    void *agcInst = WebRtcAgc_Create();    if (agcInst == NULL) return -1;    int status = WebRtcAgc_Init(agcInst, minLevel, maxLevel, agcMode, sampleRate);    if (status != 0) {        printf("WebRtcAgc_Init fail\n");        WebRtcAgc_Free(agcInst);        return -1;    }    status = WebRtcAgc_set_config(agcInst, agcConfig);    if (status != 0) {        printf("WebRtcAgc_set_config fail\n");        WebRtcAgc_Free(agcInst);        return -1;    }    size_t num_bands = 1;    int inMicLevel, outMicLevel = -1;    int16_t out_buffer[maxSamples];    int16_t *out16 = out_buffer;    uint8_t saturationWarning = 1;                 //是否有溢出發生，增益放大以後的最大值超過了65536    int16_t echo = 0;                                 //增益放大是否考慮回聲影響    for (int i = 0; i < nTotal; i++) {        inMicLevel = 0;        int nAgcRet = WebRtcAgc_Process(agcInst, (const int16_t *const *) &input, num_bands, samples,                                        (int16_t *const *) &out16, inMicLevel, &outMicLevel, echo,                                        &saturationWarning);        if (nAgcRet != 0) {            printf("failed in WebRtcAgc_Process\n");            WebRtcAgc_Free(agcInst);            return -1;        }        memcpy(input, out_buffer, samples * sizeof(int16_t));        input += samples;    }    WebRtcAgc_Free(agcInst);    return 1;}void auto_gain(char *in_file, char *out_file) {    //音頻採樣率    uint32_t sampleRate = 0;    //總音頻採樣數    uint64_t inSampleCount = 0;    int16_t *inBuffer = wavRead_int16(in_file, &sampleRate, &inSampleCount);    //如果載入成功    if (inBuffer != nullptr) {        //  kAgcModeAdaptiveAnalog  類比音量大小        //  kAgcModeAdaptiveDigital 自適應增益        //  kAgcModeFixedDigital 固定增益        agcProcess(inBuffer, sampleRate, inSampleCount, kAgcModeAdaptiveDigital);        wavWrite_int16(out_file, inBuffer, sampleRate, inSampleCount);        free(inBuffer);    }}int main(int argc, char *argv[]) {    printf("WebRTC Automatic Gain Control\n");    printf("部落格:http://cpuimage.cnblogs.com/\n");    printf("音頻自動增益\n");    if (argc < 2)        return -1;    char *in_file = argv[1];    char drive[3];    char dir[256];    char fname[256];    char ext[256];    char out_file[1024];    splitpath(in_file, drive, dir, fname, ext);    sprintf(out_file, "%s%s%s_out%s", drive, dir, fname, ext);    auto_gain(in_file, out_file);    printf("按任意鍵退出程式 \n");    getchar();    return 0;}

 

 靜音檢測完整範例程式碼：

#include <stdio.h>#include <stdlib.h>#include <stdint.h>//採用https://github.com/mackron/dr_libs/blob/master/dr_wav.h 解碼#define DR_WAV_IMPLEMENTATION#include "dr_wav.h"#include "vad.h"#ifndef nullptr#define nullptr 0#endif#ifndef MIN#define  MIN(A, B)        ((A) < (B) ? (A) : (B))#endif#ifndef MAX#define  MAX(A, B)        ((A) > (B) ? (A) : (B))#endif//讀取wav檔案int16_t *wavRead_int16(char *filename, uint32_t *sampleRate, uint64_t *totalSampleCount) {    unsigned int channels;    int16_t *buffer = drwav_open_and_read_file_s16(filename, &channels, sampleRate, totalSampleCount);    if (buffer == nullptr) {        printf("讀取wav檔案失敗.");    }    //僅僅處理單通道音頻    if (channels != 1) {        drwav_free(buffer);        buffer = nullptr;        *sampleRate = 0;        *totalSampleCount = 0;    }    return buffer;}int vadProcess(int16_t *buffer, uint32_t sampleRate, size_t samplesCount, int16_t vad_mode, int per_ms_frames) {    if (buffer == nullptr) return -1;    if (samplesCount == 0) return -1;    // kValidRates : 8000, 16000, 32000, 48000    // 10, 20 or 30 ms frames    per_ms_frames = MAX(MIN(30, per_ms_frames), 10);    size_t samples = sampleRate * per_ms_frames / 1000;    if (samples == 0) return -1;    int16_t *input = buffer;    size_t nTotal = (samplesCount / samples);    void *vadInst = WebRtcVad_Create();    if (vadInst == NULL) return -1;    int status = WebRtcVad_Init(vadInst);    if (status != 0) {        printf("WebRtcVad_Init fail\n");        WebRtcVad_Free(vadInst);        return -1;    }    status = WebRtcVad_set_mode(vadInst, vad_mode);    if (status != 0) {        printf("WebRtcVad_set_mode fail\n");        WebRtcVad_Free(vadInst);        return -1;    }    printf("Activity ： \n");    for (int i = 0; i < nTotal; i++) {        int nVadRet = WebRtcVad_Process(vadInst, sampleRate, input, samples);        if (nVadRet == -1) {            printf("failed in WebRtcVad_Process\n");            WebRtcVad_Free(vadInst);            return -1;        } else {            // output result            printf(" %d \t", nVadRet);        }        input += samples;    }    printf("\n");    WebRtcVad_Free(vadInst);    return 1;}void vad(char *in_file) {    //音頻採樣率    uint32_t sampleRate = 0;    //總音頻採樣數    uint64_t inSampleCount = 0;    int16_t *inBuffer = wavRead_int16(in_file, &sampleRate, &inSampleCount);    //如果載入成功    if (inBuffer != nullptr) {        //    Aggressiveness mode (0, 1, 2, or 3)        int16_t mode = 1;        int per_ms = 30;        vadProcess(inBuffer, sampleRate, inSampleCount, mode, per_ms);        free(inBuffer);    }}int main(int argc, char *argv[]) {    printf("WebRTC Voice Activity Detector\n");    printf("部落格:http://cpuimage.cnblogs.com/\n");    printf("靜音檢測\n");    if (argc < 2)        return -1;    char *in_file = argv[1];    vad(in_file);    printf("按任意鍵退出程式 \n");    getchar();    return 0;}

自動增益項目地址:https://github.com/cpuimage/WebRTC_AGC

具體流程為： 

載入wav(拖放wav檔案到可執行檔上)->增益處理->儲存為_out.wav檔案

 

靜音檢測項目地址：https://github.com/cpuimage/WebRTC_VAD

具體流程為： 

載入wav(拖放wav檔案到可執行檔上)->輸出靜音檢測結果

備忘 :1 為非靜音，0 為靜音

 該注意的地方和參數，見代碼注釋。

用cmake即可進行編譯範例程式碼，詳情見CMakeLists.txt。

 

若有其他相關問題或者需求也可以郵件聯絡俺探討。

郵箱地址是: 
[email protected]

音頻自動增益 與 靜音檢測 演算法 附完整C代碼【轉】