hslab.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                  */ 
/*                                                             */
/*                                                             */
/* ----------------------------------------------------------- */
/* developed at:                                               */
/*                                                             */
/*      Speech Vision and Robotics group                       */
/*      Cambridge University Engineering Department            */
/*      http://svr-www.eng.cam.ac.uk/                          */
/*                                                             */
/*      Entropic Cambridge Research Laboratory                 */
/*      (now part of Microsoft)                                */
/*                                                             */
/* ----------------------------------------------------------- */
/*         Copyright: Microsoft Corporation                    */
/*          1995-2000 Redmond, Washington USA                  */
/*                    http://www.microsoft.com                 */
/*                                                             */
/*              2001  Cambridge University                     */
/*                    Engineering Department                   */
/*                                                             */
/*   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: HSLab.c:   The Speech Label Editor            */
/* ----------------------------------------------------------- */

char *hslab_version = "!HVER!HSLab:   3.3 [CUED 28/04/05]";
char *hslab_vc_id = "$Id: HSLab.c,v 1.2 2005/05/12 15:51:29 jal58 Exp $";

/* 
   --------------------------------------------------------------
   
   In order to compile this program, you may need to use different 
   compiler options.  For example, under HP-UX 8.07 you would add 
   
   -D_HPUX_SOURCE
   
   to the compiler flags.  Also the environment variable 
   LPATH will need to be set to "/lib:/usr/lib:/usr/lib/X11R4"

   --------------------------------------------------------------
*/

/*
  Enable use of stat calls to reload files changed by another process - e.g. 
  label files generated by a recogniser

  #define USE_STAT
*/

#define USE_STAT


#include "HShell.h"
#include "HMem.h"
#include "HMath.h"
#include "HWave.h"
#include "HAudio.h"
#include "HLabel.h"
#include "HGraf.h"

#define WINNAME      "HSLab"
#define HSLAB_INFO "HSLab V2.0 by Valtcho Valtchev and Steve Young"

#define SLEN  256       /* string length for various buffers */
#define MAX_ZOOM 100       /* maximum level of zooming */

#define INIT_XPOS 10       /* initial x-position of the HSLab window */
#define INIT_YPOS 10       /* initial y-position of the HSLab window */
#define WIDTH    700       /* window width */
#define HEIGHT   500       /* window height */

#define MAX_LAB_LEN 128   /* maximum length of a string representing a label */
#define LAB_BUF_LEN 256   /* string length for buffers used to store labels */


/* define 16-bit resolution, assume that shorts are 16-bit */
#define MIN_AMPL   SHRT_MIN 
#define MAX_AMPL   SHRT_MAX
#define AMPL_RANGE USHRT_MAX

/* button configuration relative to the main window size */
#define BTN_AREA_W   1.00      /* area width as fraction of main window width */
#define BTN_AREA_H   0.25      /* area height as fraction of main window height */
#define BTN_AREA_X   0.0       /* top left corner x of button area as fraction of main win width */
#define BTN_AREA_Y   0.75      /* top left corner y of button area as fraction of main win height */
#define BTN_PER_ROW 10.0       /* how many buttons in a single row in button area */
#define BTN_PER_COL  3.0       /* how many buttons in a column in button area */
#define BTN_H_SPC    0.20      /* vertical spacing between rows of buttons as fraction of btn_width */
#define BTN_V_SPC    0.33      /* horizontal spacing between cols of buttons as fraction of btn_height */

#define SCROLL_PT       20
#define NUM_OF_SCALES    6     /* the number of different amplitude scales */
#define VOLUME_STEPS    10     /* volume control button steps */
#define TIMES_LABELLED   2     /* maximum number of times a frame can be labelled */

/* -------------------------- Global Types  ------------------------ */

typedef enum {
   WAVE_WIN, LAB_WIN, IO_WIN, NO_WIN 
} WinKind;

typedef enum {
   BT_NONE, BT_LOAD, BT_SAVE, BT_SCRLL, BT_SCRLR, 
   BT_ZOOM_IN, BT_ZOOM_OUT, BT_RESTORE, BT_PLAY, 
   BT_PLAY_VOL, BT_SCALE, BT_MARK, BT_UNMARK, 
   BT_LABSTR, BT_LABEL, BT_LABELAS, BT_LDELETE, 
   BT_LEDIT, BT_LSELECT, BT_ABOUT, BT_ADJUST, 
   BT_LABSET, BT_NEWSET,
   BT_UNDO, BT_SPCL, BT_REC, BT_PAUSE, BT_STOP, BT_QUIT
} BtnId;

typedef struct {           /* used to record different levels of zooming */
   long st, en;            /* start and end sample of the previous zoom level */
   int   regA, regB;       /* marked regions on the screen (if any) */
} ZoomRecord;

typedef enum {
   CREATE_OP, DELETE_OP, CHANGE_OP
} OpType;

typedef struct {        /* used to keep track of tha last operation carried out */
   OpType op;           /* the operation carried out */
   int lset;            /* label set */
   Label lab;           /* the label prior to change */
   LLink ptr;           /* pointer to the label last modified (if any) */
} UndoRecord; 

typedef struct{           /* the structure describing the various windows on the main window */
   int x, y, w, h, bw;    /* x-pos, y-pos, width, height, fg and bg colours */
   HColour fg, bg;
} RectWin;

/* ----------------------- Global Variables ---------------------------- */

#define STACKSIZE 100000     /* assume ~100K wave files */
static MemHeap tmpStack;     /* temporary storage with reset */
static MemHeap tmpCHeap;     /* storage allocated once */
static MemHeap wavStack;     /* storage for waveforms */
static MemHeap labStack;     /* storage for label objects */
static MemHeap audStack;

static short *data;          /* the data samples */
static long nSamples;        /* number of samples */
static HTime sampPeriod;     /* the sample period */
static Wave wave;            /* the input waveform */
static Boolean newData=FALSE;

static FileFormat ifmt=UNDEFF;      /* Label input file format */
static FileFormat ofmt=UNDEFF;      /* Label output file format */
static FileFormat wfmt=UNDEFF;      /* wave file format */

static HButton *btnList;     /* the list of buttons displayed on the HSLab window */
static RectWin waveWin;      /* the waveform window */
static RectWin fileWin;      /* the window for file names and other info */
static RectWin labWin1;      /* the label window */
static RectWin io_Win;       /* the I/O window used for messages and inputing strings */

static int trace = 0;                /* trace level */
static int nParm = 0;                /* total num params */
static ConfParam *cParm[MAXGLOBS];   /* configuration parameters */

static char labstr[LAB_BUF_LEN] = "Speech";
static HButton *labstr_btn;

static char volStr[10];      /* the string displayed in the volume button */
static short playVol = 0;    /* the volume for playing samples */
static HButton *vol_btn;     /* pointer to the volume button */

static HButton *lev_btn;     /* the level button */
static char levStr[10];      /* the level button string */

static int btn_v_spc;        /* calculated button vertical spacing */
static int btn_h_spc;        /* calculated button horizontal spacing */
static int btn_w;            /* calculated button width */
static int btn_h;            /* calculated button height */

static int *plotBuf;         /* buffer to store waveform samples to plot */

static int zoomLev;                    /* the current zoom level - ptr in the zoom record */
static ZoomRecord zoomRec[MAX_ZOOM];   /* the zoom record */

static float samplesPt;      /* samples per point of the graphics screen */
static long  T;              /* number of samples */
 
/* waveform scaling */
static float scValues[NUM_OF_SCALES] = {1.0, 2.0, 4.0, 8.0, 16.0, 32.0};
static char *scString[NUM_OF_SCALES] = {"x1", "x2", "x4", "x8", "x16", "x32"};
static short scIndex = 0;    /* initial scaling index == 1.0*/
static HButton *scale_btn;   /* pointer to the amplitude scale button */

static Boolean wavePtrOn = FALSE;   /* keeps track of the state of the waveform pointer */
static int thisWpos, lastWpos;      /* the positions of the waveform pointer */

static char labfn[SLEN];                   /* the label file name */
static char spfn[SLEN] = "noname.wav";     /* the speech (waveform) file name */
static char *ospfn = "noname.wav";         /* the speech file name on the command line */
static char *labDir = NULL;                /* directory for label files */
static char *labExt = "lab";               /* label file extension */
static long sStart, sEnd;                  /* the start and end sample currently visible on the screen */

static Transcription *trans;    /* the transcriptions */
static LabList *llist;          /* the current label list edited */
static int labSet;              /* the number of the current label list */
static int numSet;              /* number of alternative transcriptions */

static Boolean labsModified = FALSE;   /* tracks any changes made to the labels */
static Boolean newLabFile   = FALSE;   /* forcing the creation of a new (empty) label file */
static Boolean incLab       = FALSE;   /* increment global label */

static Boolean regnMarked = FALSE;     /* region marked */
static int markA, markB;               /* boundaries of a marked region */

static UndoRecord undo;                /* the undo record variable */
static Boolean undoEmpty;              /* shows the status of the record */

static HButton *spcl_btn;                 /* pointer to special button */
static char spcl_str[SLEN] = "Command";   /* special button string */

/* ---------------- Process Command Line ------------------------- */

void ReportUsage(void)
{
   printf("\nUSAGE: HSLab [options] waveformFile\n\n");
   printf(" Option                                       Default\n\n");
   printf(" -a      auto-increment global label          off\n");
   printf(" -i s    Output transcriptions to MLF s       off\n");  
   printf(" -n      Create new transcription             off\n");
   printf(" -s str  Set special button label             command\n");
   PrintStdOpts("FGILPX");
}

int main(int argc, char *argv[])
{
   char *s;

   void Initialise(void);
   void LoadFiles(void);
   void hRedrawWindow(void);
   void DecodeCommands(void);
   void CloseAudio(void);

   if(InitShell(argc,argv, hslab_version, hslab_vc_id)<SUCCESS)
      HError(1500,"HSLab: InitShell failed");

   InitMem(); InitMath(); 
   InitWave(); InitLabel();
   InitAudio(); InitGraf();

   if (!InfoPrinted() && NumArgs() == 0)
      ReportUsage();
   if (NumArgs() == 0) Exit(0);

   while (NextArg() == SWITCHARG) {
      s = GetSwtArg();
      if (strlen(s)!=1) 
         HError(1519,"HSLab: Bad switch %s; must be single letter",s);
      switch(s[0]){
      case 'a':
         incLab = TRUE;
         break;
      case 'i':
         if (NextArg()!=STRINGARG)
            HError(1519,"HSLab: Output MLF file name expected");
         if(SaveToMasterfile(GetStrArg())<SUCCESS)
            HError(1514,"HCopy: Cannot write to MLF");
         break;
      case 'n':
         newLabFile = TRUE;
         break;
      case 's':
         if (NextArg() != STRINGARG)
            HError(1519,"HSLab: Button label expected");
         strcpy(spcl_str, GetStrArg());
         break;
      case 'F':
         if (NextArg() != STRINGARG)
            HError(1519,"HSLab: HSLab: Data File format expected");
         if((wfmt = Str2Format(GetStrArg())) == ALIEN)
            HError(-1589,"HSLab: Warning ALIEN Data file format set");
         break;
      case 'G':
         if (NextArg() != STRINGARG)
            HError(1519,"HSLab: Label File format expected");
         if((ifmt = Str2Format(GetStrArg())) == ALIEN)
            HError(-1589,"HSLab: Warning ALIEN Label input file format set");
         break;
      case 'I':
         if (NextArg() != STRINGARG)
            HError(1519,"HSLab: Input MLF file name expected");
         LoadMasterFile(GetStrArg());
         break;
      case 'L':
         if (NextArg()!=STRINGARG)
            HError(1519,"HSLab: HSLab: Label file directory expected");
         labDir = GetStrArg(); break;
      case 'P':
         if (NextArg() != STRINGARG)
            HError(1519,"HSLab: Label File format expected");
         if((ofmt = Str2Format(GetStrArg())) == ALIEN)
            HError(-1589,"Warning ALIEN Label output file format set");
         break;
      case 'T':
         trace = GetChkedInt(0,65535,s); break;
      case 'X':
         if (NextArg()!=STRINGARG)
            HError(1519,"HSLab: Label file extension expected");
         labExt = GetStrArg(); break;
      default:
         HError(1519,"HSLab: Unknown switch %s",s);
      }
   }
   if (NextArg()==STRINGARG)
      strcpy(spfn, ospfn=GetStrArg());
   Initialise(); LoadFiles(); hRedrawWindow(); 
   DecodeCommands(); TermHGraf();
   Exit(0);
   return (0);          /* never reached -- make compiler happy */
}


/* ------------------------- Bar handling ------------------------- */

#define BAR_WIDTH   200
#define BAR_HEIGHT  15
#define BAR_BORDER  2

typedef struct {
   int x, y;
   HColour fg;
   float step;
} BarType;

static void InitBar(BarType *bar, int x_ofs, HColour fg, int range, char *str)
{
   bar->x = io_Win.x + x_ofs + HTextWidth(str) + 10;
   bar->y = io_Win.y + (io_Win.h - BAR_HEIGHT)/2;
   bar->fg = fg;
   bar->step = (float) (BAR_WIDTH - 2*BAR_BORDER)/(float) range;
   HSetColour(BLACK);
   HPrintf(io_Win.x+x_ofs, io_Win.y+(io_Win.h/2)+HTextHeight(str)/2, "%s", str);
   HSetGrey(30);
   HFillRectangle(bar->x, bar->y, bar->x+BAR_WIDTH, bar->y+BAR_HEIGHT);
}

static void StepBar(BarType *bar, int pos)
{
   int x0, y0, x1, y1;

   HSetGrey(30);
   HFillRectangle(bar->x, bar->y, bar->x+BAR_WIDTH, bar->y+BAR_HEIGHT);
   x0 = bar->x + BAR_BORDER;  y0 = bar->y + BAR_BORDER;
   x1 = x0 + ((int) bar->step*pos);   y1 = bar->y + BAR_HEIGHT-BAR_BORDER; 
   HSetColour(bar->fg);
   HFillRectangle(x0, y0, x1, y1); 
}

/* ------------------------- Audio Drivers -------------------------- */

#define REC_BUF_SIZE 1920000
typedef signed short SampleType;

BtnId DoPause(void);
short ClipSample(short sample, float sampleScale);

/* Record: record signal from audio device */
Wave Record(long *nSamples, HTime *sampPeriod)
{
   Wave w;
   AudioIn ain;
   float bufDur;
   HButton *btn;
   Boolean done;
   HEventRec hev;
   BarType tm, vu;
   char sbuf[256];
   long i, chunk;
   long nWaiting, nToRecord;
   SampleType *buf, smin, smax;
   long nEmpty, nTotal, nToFlush;
   SampleType *recbuf;

   ResetHeap(&tmpStack);
   recbuf=(SampleType *) New(&tmpStack,sizeof(SampleType)*REC_BUF_SIZE);
   /* initialise parameters */
   *sampPeriod = 0;
   ain = OpenAudioInput(&tmpStack, sampPeriod, 0.0, 0.0);
   i = (long) (1.0E+07 / (*sampPeriod));
   chunk = (long) (i / 20.0);
   bufDur = (float) REC_BUF_SIZE / (double) i;
   sprintf(sbuf, "rec buffer (0 - %.2fsec)", bufDur);
   InitBar(&tm, 10, RED, REC_BUF_SIZE, sbuf);
   InitBar(&vu, 400, DARK_GREEN, 65536, "VU:");
   buf = recbuf; done = FALSE;
   nEmpty = REC_BUF_SIZE; nTotal =0;
   /* start audio input */
   StartAudioInput(ain, NULLSIG);
   while (nEmpty > 0){
      nWaiting = SamplesInAudio(ain);
      nToRecord = (nWaiting > chunk) ? nWaiting : chunk;
      if (nEmpty < nToRecord)
         nToRecord = nEmpty;
      GetRawAudio(ain, nToRecord, buf);
      /* get min/max samples and update display*/
      smin = 32767; smax = -smin;         
      for (i=0; i<nToRecord; i++){
         if (buf[i]<smin) smin=buf[i];
         if (buf[i]>smax) smax=buf[i];
      }
      StepBar(&tm, nTotal);
      StepBar(&vu, smax-smin);
      buf += nToRecord;
      nEmpty -= nToRecord;
      nTotal += nToRecord;
      if(HEventsPending() > 0){       /* break loop if mouse  button pressed  */
         hev = HGetEvent(FALSE,NULL);
         if(hev.event == HMOUSEDOWN){
            btn = CheckButtonList(btnList, hev.x, hev.y);
            if(btn != NULL){
               switch(btn->id){
               case BT_REC: