speex开源库分析（前置处理）
1.简介
语音在采集和传输过程中，由于语音源的差异、信道的衰减、噪声的干扰以及远近效应，导致信号幅度相差很大。所有在语音处理之前我们需要对语音数据进行前置处理，包括预处理（AGC、VAD、回音消除）、重采样和噪声抑制。
所有的代码都是基于speex开源库，具体内容可以参考http://speex.org/。
本开发手册暂时只提供AGC的接口和测试代码，其他陆续添加进来。
2.接口说明
2.1 介绍
预处理模块包括自动增益控制、静音检测和回音消除。以下是接口函数，具体参考speex\ speex_preprocess.h。
| 函数名称 |
功能简介 |
| speex_preprocess_state_init |
创建预处理器 |
| speex_preprocess_state_destroy |
销毁预处理器 |
| speex_preprocess_run |
处理一帧数据 |
| speex_preprocess |
处理一帧数据（废弃的） |
| speex_preprocess_estimate_update |
更新预处理器 |
| speex_preprocess_ctl |
设置和读取预处理器的参数 |
2.1.1 speex_preprocess_state_init
| 函数原形 |
SpeexPreprocessState *speex_preprocess_state_init(int frame_size, int sampling_rate); |
| 功能 |
创建预处理器 |
| 参数 |
Frmae_size [in]每帧的大小（建议帧长为20ms） Sample_rate [in]采样率（支持8k、16k、44k） |
| 返回值 |
成功返回预处理器指针，失败返回NULL |
| 说明 |
加入是16k的语音数据，帧长20ms等于320个采样 |
2.1.2 speex_preprocess_state_destroy
| 函数原形 |
void speex_preprocess_state_destroy(SpeexPreprocessState *st); |
| 功能 |
销毁预处理器 |
| 参数 |
St [in]处理器指针 |
| 返回值 |
Void |
| 说明 |
|
2.1.3 speex_preprocess_run
| 函数原形 |
int speex_preprocess_run(SpeexPreprocessState *st, spx_int16_t *x); |
| 功能 |
处理一帧语音数据 |
| 参数 |
St [in]处理器指针 X [in|out]数据缓存，处理后的数据也存入该缓存中 |
| 返回值 |
如果VAD打开，返回值为1表示有语音，为0表示静音或者噪音 |
| 说明 |
|
2.1.4 speex_preprocess
| 函数原形 |
int speex_preprocess(SpeexPreprocessState *st, spx_int16_t *x, spx_int32_t *echo); |
| 功能 |
处理一帧语音数据（废弃的函数，简介调用speex_preprocess_run） |
| 参数 |
St [in]处理器指针 X [in|out]数据缓存，处理后的数据也存入该缓存中 |
| 返回值 |
|
| 说明 |
|
2.1.5 speex_preprocess_estimate_update
| 函数原形 |
void speex_preprocess_estimate_update(SpeexPreprocessState *st, spx_int16_t *x); |
| 功能 |
更新预处理器，不会计算输出语音 |
| 参数 |
St [in]处理器指针 X [in]数据缓存 |
| 返回值 |
Void |
| 说明 |
|
2.1.6 speex_preprocess_ctl
| 函数原形 |
int speex_preprocess_ctl(SpeexPreprocessState *st, int request, void *ptr); |
| 功能 |
设置预处理器的参数 |
| 参数 |
St [in]处理器指针 Request [in]参数的类型（由宏来代表不同的参数） Ptr [in|out]参数的值（设置参数时为in，获得参数参数时为out，这个由宏决定） |
| 返回值 |
成功返回0，失败返回-1（表明未知的请求参数） |
| 说明 |
以下数标识参数类型的宏 /** Set preprocessor denoiser state */ #define SPEEX_PREPROCESS_SET_DENOISE 0 /** Get preprocessor denoiser state */ #define SPEEX_PREPROCESS_GET_DENOISE 1
/** Set preprocessor Automatic Gain Control state */ #define SPEEX_PREPROCESS_SET_AGC 2 /** Get preprocessor Automatic Gain Control state */ #define SPEEX_PREPROCESS_GET_AGC 3
/** Set preprocessor Voice Activity Detection state */ #define SPEEX_PREPROCESS_SET_VAD 4 /** Get preprocessor Voice Activity Detection state */ #define SPEEX_PREPROCESS_GET_VAD 5
/** Set preprocessor Automatic Gain Control level (float) */ #define SPEEX_PREPROCESS_SET_AGC_LEVEL 6 /** Get preprocessor Automatic Gain Control level (float) */ #define SPEEX_PREPROCESS_GET_AGC_LEVEL 7
/** Set preprocessor dereverb state */ #define SPEEX_PREPROCESS_SET_DEREVERB 8 /** Get preprocessor dereverb state */ #define SPEEX_PREPROCESS_GET_DEREVERB 9
/** Set preprocessor dereverb level */ #define SPEEX_PREPROCESS_SET_DEREVERB_LEVEL 10 /** Get preprocessor dereverb level */ #define SPEEX_PREPROCESS_GET_DEREVERB_LEVEL 11
/** Set preprocessor dereverb decay */ #define SPEEX_PREPROCESS_SET_DEREVERB_DECAY 12 /** Get preprocessor dereverb decay */ #define SPEEX_PREPROCESS_GET_DEREVERB_DECAY 13
/** Set probability required for the VAD to go from silence to voice */ #define SPEEX_PREPROCESS_SET_PROB_START 14 /** Get probability required for the VAD to go from silence to voice */ #define SPEEX_PREPROCESS_GET_PROB_START 15
/** Set probability required for the VAD to stay in the voice state (integer percent) */ #define SPEEX_PREPROCESS_SET_PROB_CONTINUE 16 /** Get probability required for the VAD to stay in the voice state (integer percent) */ #define SPEEX_PREPROCESS_GET_PROB_CONTINUE 17
/** Set maximum attenuation of the noise in dB (negative number) */ #define SPEEX_PREPROCESS_SET_NOISE_SUPPRESS 18 /** Get maximum attenuation of the noise in dB (negative number) */ #define SPEEX_PREPROCESS_GET_NOISE_SUPPRESS 19
/** Set maximum attenuation of the residual echo in dB (negative number) */ #define SPEEX_PREPROCESS_SET_ECHO_SUPPRESS 20 /** Get maximum attenuation of the residual echo in dB (negative number) */ #define SPEEX_PREPROCESS_GET_ECHO_SUPPRESS 21
/** Set maximum attenuation of the residual echo in dB when near end is active (negative number) */ #define SPEEX_PREPROCESS_SET_ECHO_SUPPRESS_ACTIVE 22 /** Get maximum attenuation of the residual echo in dB when near end is active (negative number) */ #define SPEEX_PREPROCESS_GET_ECHO_SUPPRESS_ACTIVE 23
/** Set the corresponding echo canceller state so that residual echo suppression can be performed (NULL for no residual echo suppression) */ #define SPEEX_PREPROCESS_SET_ECHO_STATE 24 /** Get the corresponding echo canceller state */ #define SPEEX_PREPROCESS_GET_ECHO_STATE 25
/** Set maximal gain increase in dB/second (int32) */ #define SPEEX_PREPROCESS_SET_AGC_INCREMENT 26
/** Get maximal gain increase in dB/second (int32) */ #define SPEEX_PREPROCESS_GET_AGC_INCREMENT 27
/** Set maximal gain decrease in dB/second (int32) */ #define SPEEX_PREPROCESS_SET_AGC_DECREMENT 28
/** Get maximal gain decrease in dB/second (int32) */ #define SPEEX_PREPROCESS_GET_AGC_DECREMENT 29
/** Set maximal gain in dB (int32) */ #define SPEEX_PREPROCESS_SET_AGC_MAX_GAIN 30
/** Get maximal gain in dB (int32) */ #define SPEEX_PREPROCESS_GET_AGC_MAX_GAIN 31
/* Can't set loudness */ /** Get loudness */ #define SPEEX_PREPROCESS_GET_AGC_LOUDNESS 33
/* Can't set gain */ /** Get current gain (int32 percent) */ #define SPEEX_PREPROCESS_GET_AGC_GAIN 35
/* Can't set spectrum size */ /** Get spectrum size for power spectrum (int32) */ #define SPEEX_PREPROCESS_GET_PSD_SIZE 37
/* Can't set power spectrum */ /** Get power spectrum (int32[] of squared values) */ #define SPEEX_PREPROCESS_GET_PSD 39
/* Can't set noise size */ /** Get spectrum size for noise estimate (int32) */ #define SPEEX_PREPROCESS_GET_NOISE_PSD_SIZE 41
/* Can't set noise estimate */ /** Get noise estimate (int32[] of squared values) */ #define SPEEX_PREPROCESS_GET_NOISE_PSD 43
/* Can't set speech probability */ /** Get speech probability in last frame (int32). */ #define SPEEX_PREPROCESS_GET_PROB 45
/** Set preprocessor Automatic Gain Control level (int32) */ #define SPEEX_PREPROCESS_SET_AGC_TARGET 46 /** Get preprocessor Automatic Gain Control level (int32) */ #define SPEEX_PREPROCESS_GET_AGC_TARGET 47 |
3.实例代码
3.1 AGC
| #define NN 320 /* 语音数据为单通道、16bit、16k */ int _tmain(int argc, _TCHAR* argv[]) {
short in[NN]; int i; SpeexPreprocessState *st; int count=0; float f;
st = speex_preprocess_state_init(NN, 16000);
i=1; speex_preprocess_ctl(st, SPEEX_PREPROCESS_SET_AGC, &i); f=16000; speex_preprocess_ctl(st, SPEEX_PREPROCESS_SET_AGC_LEVEL, &f);
while (1) { int vad; fread(in, sizeof(short), NN, stdin); if (feof(stdin)) break; vad = speex_preprocess_run(st, in); //fprintf (stderr, "%d\n", vad); fwrite(in, sizeof(short), NN, stdout); count++; } speex_preprocess_state_destroy(st);
return 0; } |