音频自动增益 与 静音检测 算法 附完整C代码

前面分享过一个算法《音频增益响度分析 ReplayGain 附完整C代码示例》

主要用于评估一定长度音频的音量强度，

而分析之后，很多类似的需求，肯定是做音频增益，提高音量诸如此类做法。

不过在项目实测的时候，其实真的很难定标准，

到底在什么样的环境下，要增大音量，还是降低。

在通讯行业一般的做法就是采用静音检测，

一旦检测为静音或者噪音，则不做处理，反之通过一定的策略进行处理。

这里就涉及到两个算法，一个是静音检测，一个是音频增益。

增益其实没什么好说的，类似于数据归一化拉伸的做法。

静音检测 在WebRTC中 是采用计算GMM (Gaussian Mixture Model,高斯混合模型)进行特征提取的。

在很长一段时间里面，音频特征 有3个主要的方法，

GMM  ,Spectrogram (声谱图), MFCC 即 Mel-Frequency Cepstrum(Mel频率倒谱)

恕我直言,GMM 提取的特征，其鲁棒性 不如后两者。

也不多做介绍，感兴趣的同学，翻翻 * ,补补课。

当然在实际使用算法时，会由此延伸出来一些小技巧。

例如，用静音检测 来做音频裁剪，或者搭配音频增益做一些音频增强之类的操作。

自动增益在WebRTC 源代码文件是:analog_agc.c 和 digital_agc.c

静音检测 源代码文件是: webrtc_vad.c

这个命名，有一定的历史原因了。

经过梳理后，

增益算法为 agc.c agc.h

静音检测为 vad.c vad.h

增益算法的完整示例代码：

#include <stdio.h>
#include <stdlib.h>
#include <stdint.h>
//采用https://github.com/mackron/dr_libs/blob/master/dr_wav.h 解码
#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

//写wav文件
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 = ;
    format.sampleRate = (drwav_uint32) sampleRate;
    format.bitsPerSample = ;
    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();
        }
    }
}

//读取wav文件
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("读取wav文件失败.");
    }
    //仅仅处理单通道音频
    if (channels != ) {
        drwav_free(buffer);
        buffer = nullptr;
        *sampleRate = ;
        *totalSampleCount = ;
    }
    return buffer;
}

//分割路径函数
void splitpath(const char *path, char *drv, char *dir, char *name, char *ext) {
    const char *end;
    const char *p;
    const char *s;
    if (path[] && path[] == ':') {
        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 -;
    if (samplesCount == ) return -;
    WebRtcAgcConfig agcConfig;
    agcConfig.compressionGaindB = ; // default 9 dB
    agcConfig.limiterEnable = ; // default kAgcTrue (on)
    agcConfig.targetLevelDbfs = ; // default 3 (-3 dBOv)
    int minLevel = ;
    int maxLevel = ;
    size_t samples = MIN(, sampleRate / );
    if (samples == ) return -;
    const int maxSamples = ;
    int16_t *input = buffer;
    size_t nTotal = (samplesCount / samples);
    void *agcInst = WebRtcAgc_Create();
    if (agcInst == NULL) return -;
    int status = WebRtcAgc_Init(agcInst, minLevel, maxLevel, agcMode, sampleRate);
    if (status != ) {
        printf("WebRtcAgc_Init fail\n");
        WebRtcAgc_Free(agcInst);
        return -;
    }
    status = WebRtcAgc_set_config(agcInst, agcConfig);
    if (status != ) {
        printf("WebRtcAgc_set_config fail\n");
        WebRtcAgc_Free(agcInst);
        return -;
    }
    size_t num_bands = ;
    int inMicLevel, outMicLevel = -;
    int16_t out_buffer[maxSamples];
    int16_t *out16 = out_buffer;
    uint8_t saturationWarning = ;                 //是否有溢出发生，增益放大以后的最大值超过了65536
    int16_t echo = ;                                 //增益放大是否考虑回声影响
    for (int i = ; i < nTotal; i++) {
        inMicLevel = ;
        int nAgcRet = WebRtcAgc_Process(agcInst, (const int16_t *const *) &input, num_bands, samples,
                                        (int16_t *const *) &out16, inMicLevel, &outMicLevel, echo,
                                        &saturationWarning);

        if (nAgcRet != ) {
            printf("failed in WebRtcAgc_Process\n");
            WebRtcAgc_Free(agcInst);
            return -;
        }
        memcpy(input, out_buffer, samples * sizeof(int16_t));
        input += samples;
    }
    WebRtcAgc_Free(agcInst);
    return ;
}

void auto_gain(char *in_file, char *out_file) {
    //音频采样率
    uint32_t sampleRate = ;
    //总音频采样数
    uint64_t inSampleCount = ;
    int16_t *inBuffer = wavRead_int16(in_file, &sampleRate, &inSampleCount);
    //如果加载成功
    if (inBuffer != nullptr) {
        //  kAgcModeAdaptiveAnalog  模拟音量调节
        //  kAgcModeAdaptiveDigital 自适应增益
        //  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("博客:http://cpuimage.cnblogs.com/\n");
    printf("音频自动增益\n");
    if (argc < )
        return -;
    char *in_file = argv[];
    char drive[];
    char dir[];
    char fname[];
    char ext[];
    char out_file[];
    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("按任意键退出程序 \n");
    getchar();
    return ;
}

静音检测完整示例代码：

#include <stdio.h>
#include <stdlib.h>
#include <stdint.h>
//采用https://github.com/mackron/dr_libs/blob/master/dr_wav.h 解码
#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

//读取wav文件
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("读取wav文件失败.");
    }
    //仅仅处理单通道音频
    if (channels != ) {
        drwav_free(buffer);
        buffer = nullptr;
        *sampleRate = ;
        *totalSampleCount = ;
    }
    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 -;
    if (samplesCount == ) return -;
    // kValidRates : 8000, 16000, 32000, 48000
    // 10, 20 or 30 ms frames
    per_ms_frames = MAX(MIN(, per_ms_frames), );
    size_t samples = sampleRate * per_ms_frames / ;
    if (samples == ) return -;
    int16_t *input = buffer;
    size_t nTotal = (samplesCount / samples);

    void *vadInst = WebRtcVad_Create();
    if (vadInst == NULL) return -;
    int status = WebRtcVad_Init(vadInst);
    if (status != ) {
        printf("WebRtcVad_Init fail\n");
        WebRtcVad_Free(vadInst);
        return -;
    }
    status = WebRtcVad_set_mode(vadInst, vad_mode);
    if (status != ) {
        printf("WebRtcVad_set_mode fail\n");
        WebRtcVad_Free(vadInst);
        return -;
    }
    printf("Activity ： \n");
    for (int i = ; i < nTotal; i++) {
        int nVadRet = WebRtcVad_Process(vadInst, sampleRate, input, samples);
        if (nVadRet == -) {
            printf("failed in WebRtcVad_Process\n");
            WebRtcVad_Free(vadInst);
            return -;
        } else {
            // output result
            printf(" %d \t", nVadRet);
        }
        input += samples;
    }
    printf("\n");
    WebRtcVad_Free(vadInst);
    return ;
}

void vad(char *in_file) {
    //音频采样率
    uint32_t sampleRate = ;
    //总音频采样数
    uint64_t inSampleCount = ;
    int16_t *inBuffer = wavRead_int16(in_file, &sampleRate, &inSampleCount);
    //如果加载成功
    if (inBuffer != nullptr) {
        //    Aggressiveness mode (0, 1, 2, or 3)
        int16_t mode = ;
        int per_ms = ;
        vadProcess(inBuffer, sampleRate, inSampleCount, mode, per_ms);
        free(inBuffer);
    }
}

int main(int argc, char *argv[]) {
    printf("WebRTC Voice Activity Detector\n");
    printf("博客:http://cpuimage.cnblogs.com/\n");
    printf("静音检测\n");
    if (argc < )
        return -;
    char *in_file = argv[];
    vad(in_file);
    printf("按任意键退出程序 \n");
    getchar();
    return ;
}

自动增益项目地址:https://github.com/cpuimage/WebRTC_AGC

具体流程为：

加载wav(拖放wav文件到可执行文件上)->增益处理->保存为_out.wav文件

静音检测项目地址：https://github.com/cpuimage/WebRTC_VAD

具体流程为：

加载wav(拖放wav文件到可执行文件上)->输出静音检测结果

备注 :1 为非静音，0 为静音

该注意的地方和参数，见代码注释。

用cmake即可进行编译示例代码，详情见CMakeLists.txt。

若有其他相关问题或者需求也可以邮件联系俺探讨。

邮箱地址是: 
[e-mail]