hlrescore.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/                          */
/*                                                             */
/* author: Gunnar Evermann <[e-mail]>               */
/* ----------------------------------------------------------- */
/*         Copyright:                                          */
/*         2001-2002  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: HLRescore.c: Lattice rescoring/pruning          */
/* ----------------------------------------------------------- */

/*#### todo:

     - implement lattice oracle WER calculation
     - allow batch processing?
*/

char *hlrescore_version = "!HVER!HLRescore:   3.3 [CUED 28/04/05]";
char *hlrescore_vc_id = "$Id: HLRescore.c,v 1.1.1.1 2005/05/12 10:52:54 jal58 Exp $";

#include "HShell.h"
#include "HMem.h"
#include "HMath.h"
#include "HWave.h"
#include "HLabel.h"
#include "HAudio.h"
#include "HParm.h"
#include "HModel.h"
#include "HUtil.h"
#include "HDict.h"
#include "HNet.h"
#include "HRec.h"
#include "HLM.h"
#include "HLat.h"

/* -------------------------- Trace Flags & Vars ------------------------ */

#define T_TOP  00001      /* Basic progress reporting */
#define T_TRAN 00002      /* Output transcriptions */
#define T_LAT  00004      /* Lattice I/O */
#define T_MEM  00010      /* Memory usage, start and finish */

static int trace = 0;

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

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

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

static char *dictfn;		/* dict filename from commandline */
static char *latInDir = NULL;   /* Lattice input dir, set by -L  */
static char *latInExt = "lat";  /* Lattice Extension, set by -X */

static FileFormat ofmt=UNDEFF;  /* Label output file format */
static char *labOutDir = NULL;  /* output label file directory */
static char *labOutExt = "rec"; /* output label file extension */
static char *labOutForm = NULL; /* output label format */
static char *latOutForm = NULL; /* output lattice format */

static double lmScale = 1.0;    /* LM scale factor */
static double acScale = 1.0;    /* acoustic scale factor */
static LogDouble wordPen = 0.0; /* inter word log penalty */
static double prScale = 1.0;    /* pronunciation scale factor */

static Boolean fixPronprobs = FALSE; /* get pron probs from dict */

static char *lmFile = NULL;     /* LM filename */
static LModel *lm;              /* LM for expandin lattices */
static Vocab vocab;		/* wordlist or dictionary */

static char *startWord;         /* word at start of Lattice (!SENT_START) */
static char *endWord;           /* word at end of Lattice (!SENT_END) */
static char *startLMWord;       /* word at start in LM (<s>) */
static char *endLMWord;         /* word at end in LM (</s>) */

static Boolean fixBadLats = FALSE;         /* fix final word in lattices */
static Boolean sortLattice = TRUE;         /* sort lattice nodes by time & posterior */

static LogDouble pruneInThresh = - LZERO;  /* beam for pruning (-t) */
static LogDouble pruneOutThresh = - LZERO; /* beam for pruning (-u) */
static LogDouble pruneInArcsPerSec = 0.0;  /* arcs per second threshold (-t) */
static LogDouble pruneOutArcsPerSec = 0.0; /* arcs per second threshold (-u) */

/* operations to perform: */
static Boolean pruneInLat = FALSE;  /* -t */
static Boolean writeLat = FALSE;    /* -w */
static Boolean expandLat = FALSE;   /* -n */
static Boolean pruneOutLat = FALSE; /* -u */
static Boolean findBest = FALSE;    /* -f */
static Boolean calcStats = FALSE;   /* -c */


/* -------------------------- Heaps ------------------------------------- */

static MemHeap latHeap;
static MemHeap lmHeap;
static MemHeap transHeap;

/* -------------------------- Prototypes -------------------------------- */

void SetConfParms (void);
void ReportUsage (void);
void ProcessLattice (char *latfn);


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

/* SetConfParms: set conf parms relevant to this tool */
void SetConfParms(void)
{
   int i;
   char buf[MAXSTRLEN];
   Boolean b;

   nParm = GetConfig ("HLRESCORE", TRUE, cParm, MAXGLOBS);
   if (nParm>0){
      if (GetConfInt (cParm, nParm, "TRACE", &i)) trace = i;
      if (GetConfStr (cParm, nParm, "STARTWORD", buf))
         startWord = CopyString (&gstack, buf);
      if (GetConfStr (cParm, nParm, "ENDWORD", buf))
         endWord = CopyString (&gstack, buf);
      if (GetConfStr (cParm, nParm, "STARTLMWORD", buf))
         startLMWord = CopyString (&gstack, buf);
      if (GetConfStr (cParm, nParm, "ENDLMWORD", buf))
         endLMWord = CopyString (&gstack, buf);
      if (GetConfBool (cParm, nParm, "FIXBADLATS", &b)) fixBadLats = b;
      if (GetConfBool (cParm, nParm, "SORTLATTICE", &b)) sortLattice = b;
   }
}

void ReportUsage(void)
{
   printf("\nUSAGE: HLRescore [options] vocabFile Files...\n\n");
   printf(" Option                                   Default\n\n");
   printf(" -i s    Output transcriptions to MLF s       off\n"); 
   printf(" -l s    dir to store label/lattice files     current\n");
   printf(" -n s    load n-gram LM and expand lattice    off\n");
   printf(" -o s    output label formating NCSTWMX       none\n");
   printf(" -t f [f] input lattice pruning threshold     off\n");
   printf(" -u f [f] output lattice pruning threshold    off\n");
   printf(" -p f    inter model trans penalty (log)      0.0\n");
   printf(" -s f    grammar scale factor                 1.0\n");
   printf(" -a f    acoustic scale factor                1.0\n");
   printf(" -r f    pronunciation scale factor           1.0\n");
   printf(" -d      get pronprobs from dict              off\n");
   printf(" -c      calculate statistics                 off\n");
   printf(" -f      find 1-best transcription            off\n");
   printf(" -w      write output lattices                off\n");
   printf(" -q s    output lattice format                tvaldmnr\n"); 
   printf(" -y s    output label file extension          rec\n");
   PrintStdOpts("ILSXTGP");
   printf("\n\n");
}

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

   /*#### new error code range */
   if(InitShell (argc, argv, hlrescore_version, hlrescore_vc_id) < SUCCESS)
      HError (4000, "HLRescore: InitShell failed");
  
   InitMem();
   InitMath();
   InitWave();
   InitLabel();
   InitAudio();
   if (InitParm()<SUCCESS) 
      HError(4000,"HLRescore: InitParm failed");
   InitModel();
   InitUtil();
   InitDict();
   InitNet(); 
   InitRec();
   InitLM();
   InitLat();

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

   while (NextArg() == SWITCHARG) {
      s = GetSwtArg();
      if (strlen(s) != 1) 
         HError (4019, "HLRescore: Bad switch %s; must be single letter",s);
      switch (s[0]){
      case 'L':
         if (NextArg() != STRINGARG)
            HError (4019,"HLRescore: Lattice file directory expected");
         latInDir = GetStrArg(); 
         break;
      case 'X':
         if (NextArg() != STRINGARG)
            HError (4019, "HLRescore: Lattice filename extension expected");
         latInExt = GetStrArg(); 
         break;
      case 'i':
         if (NextArg() != STRINGARG)
            HError (4019, "HLRescore: Output MLF file name expected");
         if (SaveToMasterfile (GetStrArg()) < SUCCESS)
            HError (4014, "HLRescore: Cannot write to MLF");
         break;

      case 'I':
         if (NextArg() != STRINGARG)
            HError (4019, "HLRescore: Input MLF file name expected");
         LoadMasterFile (GetStrArg ());
         break;

      case 'P':
         if (NextArg() != STRINGARG)
            HError (4019, "HLRescore: Output Label File format expected");
         if((ofmt = Str2Format (GetStrArg ())) == ALIEN)
            HError (-4089, "HLRescore: Warning ALIEN Label output file format set");
         break;
      case 'q':
	 if (NextArg () != STRINGARG)
	    HError (4019, "HLRescore: Output lattice format expected");
	 latOutForm = GetStrArg ();
	 break;
      
      case 'l':
         if (NextArg() != STRINGARG)
            HError (4019, "HLRescore: Output Label file directory expected");
         labOutDir = GetStrArg(); 
         break;
      case 'o':
         if (NextArg() != STRINGARG)
            HError (4019, "HLRescore: Output label format expected");
         labOutForm = GetStrArg(); 
         break;
      case 'y':
         if (NextArg() != STRINGARG)
            HError (4019, "HLRescore: Output label file extension expected");
         labOutExt = GetStrArg(); break;