使用speex对pcm,wav进行降噪处理

1. speex的降噪模块的简介

speex的语音处理模块要使用独立于 speex codec库的libspeexdsp 库。

这个分离的库是在1.2版本后实现；

它这库包括了： 预处理，回声消除，jitter buffer 和重采样模块；

在Unix/Linux环境下，使用 -lspeexdsp -lm 来编译和链接。

 

和libspeex一样，库libspeexdsp的库函数都是可重入函数。

但它不是线程安全的，所以在多个线程中使用同一个实例时，必须加个线程安全锁。

 

NOTE:

所谓可重入函数就是允许被递归调用的函数。

函数的递归调用是 指当一个函数正被调用尚未返回时，又直接或间接调用函数本身。

一般的函数不能做到这样，只有重入函数才允许递归调用.

 

2. 下载与编译

https://www.speex.org/downloads/

http://downloads.xiph.org/releases/speex/speexdsp-1.2rc3.tar.gz

$ tar -zxvf speexdsp-1.2rc3.tar.gz

$ cd speexdsp-1.2rc3

$./configure --prefix=/data/speexdsp-1.2rc3-install/ --enable-static --disable-shared --with-pic

$ make && make install

 

3. API简介

预处理模块需要添加头文件：

#include

 

创建实例:

SpeexPreprocessState *preprocess_state = speex_preprocess_state_init(frame_size,sampling_rate);

参数： frame_size ， 建议设置成编码器相同的值。

 

对于每一个输入帧，调用处理的函数：

speex_preprocess_run(preprocess_state, audio_frame);

参数： audio_frame ， 即是输入，也是输出。

 

 

在某些场景下，有些降噪的数据并不想输出，可以下面的API:

它会更新处理器内部的状态，但不会将降噪后的数据输出，这样可以节省一些计算量：

speex_preprocess_estimate_update(preprocess_state, audio_frame);

 

 

使用下面API来改变进行预处理器的状态设置，更多的参数见下面的一节：

speex_preprocess_ctl(preprocess_state, request, ptr);

 

实例销毁 :

speex_preprocess_state_destroy(preprocess_state);

 

3.1 Preprocessor options

As with the codec, the preprocessor also has options that can be controlled using an ioctl()-like call. The available options are:

SPEEX_PREPROCESS_SET_DENOISE Turns denoising on(1) or off(2) (spx_int32_t)

SPEEX_PREPROCESS_GET_DENOISE Get denoising status (spx_int32_t)

SPEEX_PREPROCESS_SET_AGC Turns automatic gain control (AGC) on(1) or off(2) (spx_int32_t)

SPEEX_PREPROCESS_GET_AGC Get AGC status (spx_int32_t)

SPEEX_PREPROCESS_SET_VAD Turns voice activity detector (VAD) on(1) or off(2) (spx_int32_t)

SPEEX_PREPROCESS_GET_VAD Get VAD status (spx_int32_t)

SPEEX_PREPROCESS_SET_AGC_LEVEL

SPEEX_PREPROCESS_GET_AGC_LEVEL

SPEEX_PREPROCESS_SET_DEREVERB Turns reverberation removal on(1) or off(2) (spx_int32_t)

SPEEX_PREPROCESS_GET_DEREVERB Get reverberation removal status (spx_int32_t)

SPEEX_PREPROCESS_SET_DEREVERB_LEVEL Not working yet, do not use

SPEEX_PREPROCESS_GET_DEREVERB_LEVEL Not working yet, do not use

SPEEX_PREPROCESS_SET_DEREVERB_DECAY Not working yet, do not use

SPEEX_PREPROCESS_GET_DEREVERB_DECAY Not working yet, do not use

SPEEX_PREPROCESS_SET_PROB_START

SPEEX_PREPROCESS_GET_PROB_START

SPEEX_PREPROCESS_SET_PROB_CONTINUE

SPEEX_PREPROCESS_GET_PROB_CONTINUE

SPEEX_PREPROCESS_SET_NOISE_SUPPRESS Setmaximumattenuation of the noise in dB (negativespx_int32_t)

SPEEX_PREPROCESS_GET_NOISE_SUPPRESS Getmaximumattenuation of the noise in dB (negativespx_int32_t)

SPEEX_PREPROCESS_SET_ECHO_SUPPRESS Setmaximumattenuation of the residual echo in dB (negative spx_int32_t)

SPEEX_PREPROCESS_GET_ECHO_SUPPRESS Setmaximumattenuation of the residual echo in dB (negativespx_int32_t)

SPEEX_PREPROCESS_SET_ECHO_SUPPRESS_ACTIVE Set maximum attenuation of the echo in dB when near

end is active (negative spx_int32_t)

SPEEX_PREPROCESS_GET_ECHO_SUPPRESS_ACTIVE Set maximum attenuation of the echo in dB when near

end is active (negative spx_int32_t)

SPEEX_PREPROCESS_SET_ECHO_STATE Set the associated echo canceller for residual echo suppression (pointer

or NULL for no residual echo suppression)

SPEEX_PREPROCESS_GET_ECHO_STATE Get the associated echo canceller (pointer)

 

4. 应用实例

C语言实现的音频降噪代码如下。

代码中采样率、音频帧大小需要根据实际情况设置，

HEADLEN是WAV格式的文件头，占44个字节，这44个字节是不需要处理的，不然文件头会损坏，

导致得到的结果无法播放。

如果是PCM数据，则没有这个头，直接输入指定长度的数据就行；

 

noiseSuppress的值可以控制减除的噪声强度，负值越小，噪声去除的强度越大，

同时会造成原声的失真，需要作出权衡。

#include   
#include   
#include   
#include   
#include 
#include 

#define HEADLEN 44
#define SAMPLE_RATE   (48000)  
#define SAMPLES_PER_FRAME  (1024)
#define FRAME_SIZE   (SAMPLES_PER_FRAME * 1000/ SAMPLE_RATE)
#define FRAME_BYTES  (SAMPLES_PER_FRAME)
int main()
{
    size_t n = 0;
    FILE *inFile, *outFile;
    fopen_s(&inFile, "./audio/input01L.wav", "rb");
    fopen_s(&outFile, "./audio/output01L.wav", "wb");

    char *headBuf = (char*)malloc(HEADLEN);
    char *dataBuf = (char*)malloc(FRAME_BYTES * 2 );
    memset(headBuf, 0, HEADLEN);
    memset(dataBuf, 0, FRAME_BYTES);
    assert(headBuf != NULL);
    assert(dataBuf != NULL);

    SpeexPreprocessState *state = speex_preprocess_state_init(1024, SAMPLE_RATE);
    int denoise = 1;
    int noiseSuppress = -25;
    speex_preprocess_ctl(state, SPEEX_PREPROCESS_SET_DENOISE, &denoise);
    speex_preprocess_ctl(state, SPEEX_PREPROCESS_SET_NOISE_SUPPRESS, &noiseSuppress);
    
    int i;
    i = 0;
    speex_preprocess_ctl(state, SPEEX_PREPROCESS_SET_AGC, &i);
    i = 80000;
    speex_preprocess_ctl(state, SPEEX_PREPROCESS_SET_AGC_LEVEL, &i);
    i = 0;
    speex_preprocess_ctl(state, SPEEX_PREPROCESS_SET_DEREVERB, &i);
    float f = 0;
    speex_preprocess_ctl(state, SPEEX_PREPROCESS_SET_DEREVERB_DECAY, &f);
    f = 0;
    speex_preprocess_ctl(state, SPEEX_PREPROCESS_SET_DEREVERB_LEVEL, &f);

    //静音检测，效果一般
    /*
    int vad = 1;
    int vadProbStart = 80;
    int vadProbContinue = 65;
    speex_preprocess_ctl(state, SPEEX_PREPROCESS_SET_VAD, &vad); 
    speex_preprocess_ctl(state, SPEEX_PREPROCESS_SET_PROB_START, &vadProbStart); 
    speex_preprocess_ctl(state, SPEEX_PREPROCESS_SET_PROB_CONTINUE, &vadProbContinue);
    */

    bool flag = true;

    while (1)
    {
        if (flag == true)
        {
            flag = false;
            n = fread(headBuf, 1, HEADLEN, inFile);
            if (n == 0)
                break;
            fwrite(headBuf, 1, HEADLEN, outFile);
        }
        else
        {
            n = fread(dataBuf, 1, SAMPLES_PER_FRAME, inFile);
            if (n == 0)
                break;
            speex_preprocess_run(state, (spx_int16_t*)(dataBuf));
            fwrite(dataBuf, 1, SAMPLES_PER_FRAME, outFile);
        }
    }

    free(headBuf);
    free(dataBuf);
    fclose(inFile);
    fclose(outFile);
    speex_preprocess_state_destroy(state);
    return 0;
}