fe_common.h

这是一个语音特征提取的程序源码

///////////////////////////////////////////////////////////////////////////////
// This is a part of the Feature program.
// Version: 1.0
// Date: February 22, 2003
// Programmer: Oh-Wook Kwon
// Copyright(c) 2003 Oh-Wook Kwon. All rights reserved. [e-mail]
///////////////////////////////////////////////////////////////////////////////

#ifndef _FE_COMMON_H_
#define _FE_COMMON_H_

#ifdef WIN32
#include <windows.h>
#endif

#ifdef min
#undef min
#endif
#ifdef max
#undef max
#endif

#include <assert.h>
#include <iostream>
#include <cstdio>
#include <cstdlib>
#include <cstring>
#include <cmath>
#include <climits>
#include <vector>
#include <valarray>
#include <string>
using namespace std;
#pragma warning(disable:4786)


// FE_vector.h should be placed after "using namespace std;"
#include "FE_vector.h"
#include "FE_matrix.h"


// Include definition for kWaves interface
#ifdef MAIN_PROGRAM
#include "FE_wave_feature.h"
#else
#include "../kWaves-0.95/Sources/WaveFeature.h"
#endif


#ifndef FLT_MAX
#define FLT_MAX 1E37
#endif
#define LOG(x)       (((x)>0.) ? (float)log((double)(x)) : (float)(-FLT_MAX))
#define EXP(x)       (exp(x))
#define LOG10(x)     (((x)>0.) ? (float)log10((double)(x)) : (float)(-FLT_MAX))
#define EXP10(x)     ((float)pow((double)10.0,(double)(x)))


#ifndef M_PI
#define	M_PI		3.14159265358979323846
#endif
#ifndef TRUE
#define TRUE 1
#endif
#ifndef FALSE
#define FALSE 0
#endif


/*----------------------------------*/
/* Common macros                    */
/*----------------------------------*/
#ifndef my_max
#define my_max(a,b) (((a)>(b)) ? (a) : (b))
#endif
#ifndef my_min
#define my_min(a,b) (((a)<(b)) ? (a) : (b))
#endif
#ifndef my_abs
#define my_abs(x)   ((x>=0) ? (x) : (-(x)))
#endif


/*----------------------------------*/
/* Definition for feature extractor */
/*----------------------------------*/
#define SAMPLING_FREQ_1           8  /* 8kHz */
#define SAMPLING_FREQ_2          11  /* 11kHz */
#define SAMPLING_FREQ_3          16  /* 16kHz */

#define FRAME_LENGTH_1          200  /* 25ms */
#define FRAME_LENGTH_2          256  /* 23.27ms */
#define FRAME_LENGTH_3          400  /* 25ms */

#define FRAME_SHIFT_1            80  /* 10ms */
#define FRAME_SHIFT_2           110  /* 10ms */
#define FRAME_SHIFT_3           160  /* 10ms */

#define FFT_LENGTH_1            256
#define FFT_LENGTH_2            256
#define FFT_LENGTH_3            512


/************************************************************************
	User-specific Constant Values
************************************************************************/

#define DEFAULT_SAMPLING_RATE      16000 /* 16 kHz sampling */
#define DEFAULT_SHIFT_SIZE_IN_MS      10 /* 10 ms */
#define DEFAULT_WINDOW_SIZE_IN_MS     25 /* 25 ms */
#define	DEFAULT_FFT_SIZE             512 /* FFT size */


/**************************************************************
 * 	default setting
 **************************************************************/

// Preemphasis
#define FE_PRE_EMPH_FAC      0.97 /* Preemphasis factor	*/

// LPC
#define	FE_LPC_ORDER		   14 /* default LPC order */
#define	FE_LPC_ORDER_1		   12 /* LPC order for 8 kHz sampling rate */
#define	FE_LPC_ORDER_2		   14 /* LPC order for 16 kHz sampling rate	*/

// Cepstrum
#define	FE_CEP_ORDER           12 /* Number of cepstral coefficients is CEPORDER+1 including c0 */

// Dynamic features
#define FE_DELTA                5 /* Order of delta coefficients */
#define FE_DDELTA               3 /* Order of acceleration (delta-delta) coefficients */

// Filter bank
#define FE_NUM_CHANNELS        23 /* Number of bands in filter bank */
#define FE_FBANK_FLOOR    (-50.0) /* Energy floor for filter bank coefficients */


// Energy
#define FE_MIN_ENERGY (1)
#define FE_MIN_LOG_ENERGY (0)

/*----------------------------------*/
/* Data structures                  */
/*----------------------------------*/

#ifdef FeReturnCode
#define FeReturnCode __FeReturnCode
#endif

enum FeReturnCode {    /* A starting number was added to avoid overlap with other enumerators. */
	    FE_NULL=100,           /* for null transition. non-speech or non-silence */
		FE_OK,                 /* successful, no error */
		FE_WAITING,            /* waiting for speech start */
		FE_SPEECH,             /* in the middle of speech */
		FE_PAUSE,              /* pause */
		FE_END_POINT,          /* endpoint detected */
		FE_NO_SPEECH,          /* no speech found */
		FE_EARLY_END,          /* end-of-file detected before endpoint detection */
		FE_EOF,                /* end-of-file detected */
		FE_AUDIO_OPEN_ERROR,   /* error in opening audio device */
		FE_WAVE_OPEN_ERROR,    /* error in opening wave file */
		FE_STOP,               /* stopped */
		FE_UNK_ERROR           /* unknown error */
};


#endif