haudio.c

该压缩包为最新版htk的源代码,htk是现在比较流行的语音处理软件,请有兴趣的朋友下载使用

/* ----------------------------------------------------------- */
/*                                                             */
/*                          ___                                */
/*                       |_| | |_/   SPEECH                    */
/*                       | | | | \   RECOGNITION               */
/*                       =========   SOFTWARE                  */ 
/*                                                             */
/*                                                             */
/* ----------------------------------------------------------- */
/*         Copyright: Microsoft Corporation                    */
/*          1995-2000 Redmond, Washington USA                  */
/*                    http://www.microsoft.com                 */
/*                                                             */
/*   Use of this software is governed by a License Agreement   */
/*    ** See the file License for the Conditions of Use  **    */
/*    **     This banner notice must not be removed      **    */
/*                                                             */
/* ----------------------------------------------------------- */
/*      File: HAudio.c: Audio Input/Output Interface           */
/* ----------------------------------------------------------- */

char *haudio_version = "!HVER!HAudio:   3.3 [CUED 28/04/05]";
char *haudio_vc_id = "$Id: HAudio.c,v 1.2 2005/05/12 15:51:23 jal58 Exp $";

#include "HShell.h"        /* HTK Libraries */
#include "HMem.h"
#include "HMath.h"
#include "HWave.h"
#include "HAudio.h"

/* ----------------------------- Trace Flags ------------------------- */

static int trace = 0;
#define T_TOP  0001     /* Top Level tracing */
#define T_STC  0002     /* Trace Audio State Changes */
#define T_DET  0004     /* Trace Detector State Changes */
#define T_AUD  0010     /* Trace device dependent audio code */
#define T_RUN  0020     /* Trace audio read/status */

/* -------------------- Configuration Parameters --------------------- */

static ConfParam *cParm[MAXGLOBS];       /* config parameters */
static int numParm = 0;

/* ---------------------------------------------------------- */
/* 
   The type of audio device must be defined  - choices are
      NO_AUDIO    - none
      SGI_AUDIO   - SGI Indigo
      SUN16_AUDIO - Sun 16-bit linear
*/

#ifdef USS_AUDIO
#define OSS_AUDIO
#endif

#if !defined NO_AUDIO && !defined SGI_AUDIO && !defined SUN16_AUDIO && !defined SOLARIS16_AUDIO && !defined HPUX_AUDIO && !defined RS6000_AUDIO && !defined WIN95_AUDIO && !defined OSS_AUDIO && !defined HPRAW_AUDIO && !defined DEC_AUDIO && !defined WIN32_AUDIO && !defined EXT_FD_AUDIO 
#define NO_AUDIO
#endif

#ifdef WIN32_AUDIO
#define MMAPI_AUDIO
#include <windows.h>
#include <mmsystem.h>
#include <stdlib.h>    
#include <search.h>    
#endif
#ifdef DEC_AUDIO
#define MMAPI_AUDIO
#include <mme/mme_api.h>
#endif
#ifdef RS6000_AUDIO
#include <string.h>
#include <fcntl.h>
#include <UMSBAUDDevice.h>
/* function return codes */
#define rOK UMSAudioDevice_Success  
#define rUR UMSAudioDevice_UnderRun
#define rOR UMSAudioDevice_OverRun
#ifdef AIX_3_2_5
#define AUDIO_DEV "/dev/baud0"
#else
#define AUDIO_DEV "/dev/paud0"
#endif
#endif
#ifdef SGI_AUDIO
#include <audio.h>
#include <unistd.h>
#endif
#ifdef SUN16_AUDIO
#include <fcntl.h>
#include <stropts.h>
#include <sys/filio.h>
#include <sun/audioio.h>  /* May be in sys in local configuration */
#include <sys/ioctl.h>
#define AUDIO_IODEV  "/dev/audioctl"
#define AUDIO_IO     "/dev/audio"
#endif
#ifdef SOLARIS16_AUDIO
#include <fcntl.h>
#include <stropts.h>
#include <sys/filio.h>
#include <sys/audioio.h>  /* May be in sys in local configuration */
#include <sys/ioctl.h>
#define AUDIO_IODEV  "/dev/audioctl"
#define AUDIO_IO     "/dev/audio"
#define SUN16_AUDIO
#endif
#ifdef HPUX_AUDIO
#include <fcntl.h>
#include <sys/socket.h>
#include <time.h>
#include <sys/audio.h>
/* to include audio library compile with:
      -I/usr/include/audio for HPUX-9.x 
      -I/opt/audio/include for HPUX-10.x */
#include <Alib.h>
static Audio *audio_dev=NULL;
static int audio_cnt=0;
#endif
#ifdef OSS_AUDIO
#include <unistd.h>
#include <fcntl.h>
#include <sys/ioctl.h>
#include <sys/soundcard.h>
#define AUDIO_DEV "/dev/dsp"
#define MIXER_DEV "/dev/mixer"
#define AUDIO_RD 0x01
#define AUDIO_WR 0x02
#define BUF_SIZE 4096
static short zero_buf[BUF_SIZE];
static int audio_fd;
static int mixer_fd;
static unsigned short audio_io = 0x00;
static int frag_size;
static audio_buf_info audio_info;
#endif

/* Define the Audio Stream Information Records */
#define AUDBUFSIZE   640000     /* size of audio buffer */

typedef enum {
   v_peak=1,
   v_rms
} VolType;

static VolType volType = v_peak;
static Boolean lineOut = TRUE;
static Boolean phonesOut = TRUE;
static Boolean speakerOut = FALSE;
static Boolean lineIn = TRUE;
static Boolean micIn = FALSE;

static volatile Boolean stopSignalled;

typedef enum { ADS_INIT, ADS_OPEN, ADS_SAMPLING, 
               ADS_STOPPED, ADS_CLOSED } AudioDevStatus;

typedef struct {       /* circular buffer */
   Boolean isActive;     /* true if in use */
   short *data;          /* actual data buffer */
   int inx,outx;         /* in/out indices - wrap modulo size */
   int used,size;        /* used in data, size of data */
}ReplayBuf;


#ifdef MMAPI_AUDIO
#define MMAPI_BUFFER_DURATION 0.2
#define MMAPI_BUFFER_COUNT 12
static DWORD sMagic=-1;
typedef struct mmapibuf{
   int index;             /* Index of buffer */
   int size;              /* Size of buffer */
   int n;                 /* Number of valid samples in buffer */
   int cur;               /* Current sample index */
   LPWAVEHDR waveHdr;     /* Pointer to WAVEHDR */
   LPSTR waveData;        /* Data in buffer */
   struct mmapibuf *next;
   struct mmapibuf *prev;
} mmApiBuf;
#endif

typedef struct _AudioIn {
   /* -- Machine Independent Part -- */
   MemHeap *mem;             /* memory heap for this audio rec */
   HTime sampPeriod;         /* sampling period in 100ns units */
   int frSize;               /* num samples per speech frame */
   int frRate;               /* num samples between speech frames */
   short * frBuf;            /* buffer for constructing frames */
   Vector frOLap;            /* frame overlap buffer used by GetAudio */
   int inOLap;               /* num samples in frOLap */
   AudioInStatus status;     /* current status of this audio stream */
   short buffer[AUDBUFSIZE];
   int bufferSize;           /* Size of audio buffer */
   int nInBuffer;            /* Number of valid samples in buffer */
   int inBufPos;             /* Position to write in to buffer */
   int outBufPos;            /* Position to read out of buffer */
   int sig;                  /* signal if any */
   ReplayBuf rbuf;           /* replay buffer (if needed) */
   AudioDevStatus isActive;  /* indicates when device active */
   float curVol;             /* Current volume of input speech */
   /* -- Machine Dependent Part -- */
#ifdef MMAPI_AUDIO
   MMRESULT mmError;
   DWORD magic;             /* Magic number identifying this instance */
   HWAVEIN waveIn;
   LPPCMWAVEFORMAT waveFmt; /* Pointer to PCMWAVEFORMAT */
   LPMMTIME wavePos;        /* Pointer to MMTIME */
   int total;               /* Total number of samples queued */
   int current;             /* Index of current buffer */
   int bufSize;             /* Block size for each buffer */
   mmApiBuf *qHead;         /* Head of buffer waiting to be filled list */
   mmApiBuf *qTail;         /* Tail of buffer waiting to be filled list */
   mmApiBuf *fHead;         /* Head of filled buffer list */
   mmApiBuf *fTail;         /* Tail of filled buffer list */
#ifdef WIN32_AUDIO
   CRITICAL_SECTION c;
   HANDLE callBackEvent;
#endif
#endif
#ifdef RS6000_AUDIO
   UMSBAUDDevice adevin;
   UMSAudioDevice_ReturnCode rc;
   Environment *evin;
   long sw;
   long osamples;            /* Sample rate */
   char *obyte_order;
   long lgain;
   long rgain;
   long channels ;
   long bits ;
   char inConn[30];   
#endif
#ifdef SGI_AUDIO
   long params[2];           /* parameter array */
   ALport recPort;           /* SGI audio port */
   ALconfig config;          /* configuration record */
#endif
#ifdef SUN16_AUDIO
   int numSamples;
   int audio_ctld;
   audio_info_t audio_info;
#endif
#ifdef HPUX_AUDIO
   Audio *audio;
   ATransID tid;
   AudioAttributes attr;
   SSRecordParams parms;
   AGainEntry gains[4];
   SStream stream;
   int socket;
#endif
}AudioInRec;

typedef struct _AudioOut {
   /* -- Machine Independent Part -- */
   MemHeap *mem;             /* memory heap for this audio rec */
   float vol;                /* current volume */
   Boolean isActive;         /* true when device active */
   /* -- Machine Dependent Part -- */
#ifdef MMAPI_AUDIO
   MMRESULT mmError;
   DWORD magic;             /* Magic number identifying this instance */
   HWAVEOUT waveOut;
   UINT waveOutDev;         /* Device being used for wave output */
   LPPCMWAVEFORMAT waveFmt; /* Pointer to PCMWAVEFORMAT */
   LPMMTIME wavePos;        /* Pointer to MMTIME */
   int total;               /* Total number of samples queued */
   int current;             /* Index of current buffer */
   mmApiBuf *pHead;         /* Head of buffer list */
   mmApiBuf *pTail;         /* Tail of buffer list */
#ifdef WIN32_AUDIO
   CRITICAL_SECTION c;
   HANDLE callBackEvent;
#endif
#endif
#ifdef RS6000_AUDIO
   UMSBAUDDevice adevout;
   UMSAudioDevice_ReturnCode rc;
   Environment *evout;
   long sw;
   long osamples;            /* Sample rate */
   char *obyte_order;
   long lgain;
   long rgain;
   long channels ;
   long bits ;
   char outConn[30];   
#endif
#ifdef SGI_AUDIO
   long params[6];           /* parameter array */
   ALport playPort;          /* SGI audio port */
   ALconfig config;          /* configuration record */
#endif
#ifdef SUN16_AUDIO
   int numSamples;
   int numWrites;
   int audio_ctld;
   audio_info_t audio_info;
#endif
#ifdef HPUX_AUDIO
   Audio *audio;
   ATransID tid;
   int nToPlay;
   AudioAttributes attr;
   SSPlayParams parms;
   AGainEntry gains[4];
   SStream stream;
   int socket;
#endif
}AudioOutRec;

/* ------------------ Device Dependent Routines ----------------- */
/* All device dependent parts of this module are in this section  */
/* -------------------------------------------------------------- */

#ifdef SUN16_AUDIO
#define NUM_SAMP_FREQS 10
static float sampFreqs[NUM_SAMP_FREQS] = {
   8000, 9600, 11025, 16000, 18900, 22050, 32000, 37800, 44100, 48000
};

/* TrimSampFreq: find the nearest available sampling frequency */
static int TrimSampFreq(int f)
{
   int i, d;
   int min, mi;
  
   min = (int) fabs((double)(sampFreqs[0]-f)); mi = 0;
   for (i=0; i<NUM_SAMP_FREQS; i++) {
      d = (int) fabs((double)(sampFreqs[i]-f));
      if (d < min) {
         min = d; mi = i;
      }
   }
   return (int) sampFreqs[mi];
}
#endif

#ifdef OSS_AUDIO
/* IsVAXOrder: returns true if machine has VAX ordered bytes */
static Boolean IsVAXOrder(void)
{
   short x, *px;
   unsigned char *pc;
   
   px = &x;
   pc = (unsigned char *) px;
   *pc = 1; *(pc+1) = 0;         /* store bytes 1 0 */
   return x==1;          /* does it read back as 1? */
}
#endif

/* CalcVolume: calculate volume of data */
static float CalcVolume(short *data, int len)
{
   float vol;
   double sum, sqr;
   int i, minSamp, maxSamp;
   
   switch (volType) {
   case v_rms:
      sum=sqr=0.0;
      for(i = 0; i < len; i++) {
         sum += (double)data[i];
         sqr += ((double)data[i]) * ((double)data[i]);
      }
      sum/=len; sqr/=len;
      vol = sqrt(sqr-sum*sum);
      break;
   case v_peak:
   default:
      minSamp =maxSamp = data[0];
      for(i = 0; i < len; i++) {
         if ( data[i] > maxSamp )  maxSamp = data[i];
         else if ( data[i] < minSamp )  minSamp = data[i];
      }
      vol = (maxSamp-minSamp)/2.0;
      break;
   }
   return vol;
}


#ifdef MMAPI_AUDIO
#ifdef WIN32_AUDIO
void *mmeAllocMem(size_t size)
{
   void *ptr;
   ptr=GlobalAlloc(GMEM_FIXED,size);
   if (ptr==NULL)
      HError(6006,"StartAudi: Cannot allocate memory for mme structure");
   return(ptr);
}

void *mmeAllocBuffer(size_t size)
{
   void *ptr;
   ptr=GlobalAlloc(GMEM_FIXED,size);
   if (ptr==NULL)
      HError(6006,"StartAudi: Cannot allocate memory for mme structure");
   return(ptr);
}

Boolean mmeFreeMem(void *ptr)
{
   ptr=GlobalFree(ptr);
   return(TRUE);
}

Boolean mmeFreeBuffer(void *ptr)
{