headapt.c

隐马尔科夫模型工具箱

/* ----------------------------------------------------------- */
/*                                                             */
/*                          ___                                */
/*                       |_| | |_/   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                 */
/*                                                             */
/*              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: HEAdapt.c: Adaptation Tool                */
/* ----------------------------------------------------------- */

char *headapt_version = "!HVER!HEAdapt:   3.2 [CUED 09/12/02]";
char *headapt_vc_id = "$Id: HEAdapt.c,v 1.10 2002/12/19 16:37:40 ge204 Exp $";

#include "HShell.h"     /* HMM ToolKit Modules */
#include "HMem.h"
#include "HMath.h"
#include "HSigP.h"
#include "HAudio.h"
#include "HWave.h"
#include "HVQ.h"
#include "HParm.h"
#include "HLabel.h"
#include "HModel.h"
#include "HTrain.h"
#include "HUtil.h"
#include "HAdapt.h"
#include "HFB.h"


/* Trace Flags */
#define T_TOP   0001    /* Top level tracing */
#define T_MAP   0002    /* logical/physical hmm map */

Boolean traceHFB = FALSE;        /* pass to HFB to retain top-level tracing */

/* Global Settings */

static char * labDir = NULL;      /* label (transcription) file directory */
static char * labExt = "lab";     /* label file extension */
static char * hmmDir = NULL;      /* directory to look for hmm def files */
static char * hmmExt = NULL;      /* hmm def file extension */
static char * newDir = NULL;      /* directory to store new hmm def files */
static char * newExt = NULL;      /* extension of new reestimated hmm files */
static char * saveTransFN=NULL;   /* stats file, if any */
static char * uid ="unknown";     /* temporary uid */
static char * uname ="Unknown";   /* temporary uname */
static char * chan ="Standard";   /* temporary mic channel name */
static char * desc ="None";       /* temporary description to save in tmf */
static char * transFile = NULL;   /* file holding transformations */
static int blocks=0;              /* number of block diagonal matrices */
static RegClassType regClass=ADPTUNDEF;/* regClass type to be used */
static RegTransType regTrans=TRANSUNDEF;/* reg transform type to be used */

static float minVar  = 0.0;       /* minimum variance (diagonal only) */
static float occThresh=0.0;       /* Regression tree ocupation count minimum */
static int update    = 0;         /* update trans every "update" utterances */

static int trace     = 0;         /* Trace level */
static Boolean saveBinary=FALSE;  /* save output in binary  */
static FileFormat dff=UNDEFF;     /* data file format */
static FileFormat lff=UNDEFF;     /* label file format */

/* Global Data Structures - valid for all training utterances */
static LogDouble pruneInit = NOPRUNE;    /* pruning threshold initially */
static LogDouble pruneInc = 0.0;         /* pruning threshold increment */
static LogDouble pruneLim = NOPRUNE;     /* pruning threshold limit */
static float minFrwdP = 10.0;    /* mix prune threshold */

static float tau         = 15.0;    /* scaling for MAP adaptation */
static Boolean map       = FALSE;   /* do MAP adaptation */
static Boolean mapUseMLLR= FALSE;   /* do MAP + MLLR adaptation */
static Boolean global    = FALSE;   /* do global adaptation only */

static Boolean firstTime = TRUE;    /* Flag used to enable creation of ot */
static Boolean twoDataFiles = FALSE; /* Enables creation of ot2 for FB
                                        training using two data files */
static Vector vFloor[SMAX]; /* variance floor - default is all zero */

static LogDouble totalPr=0.0;   /* total log prob upto current utterance */
static int totalT=0;            /* total number of frames in training data */

static MemHeap hmmStack;   /*For Storage of all dynamic structures created...*/
static MemHeap uttStack;
static MemHeap regStack;
static MemHeap fbInfoStack;

/* ------------------ Process Command Line -------------------------- */
   
/* SetConfParms: set conf parms relevant to HCompV  */
void SetConfParms(void)
{
   int i;
   Boolean b;
   ConfParam *cParm[MAXGLOBS];/* configuration parameters */
   int nParm = 0;             /* total num params */
   
   nParm = GetConfig("HEADAPT", TRUE, cParm, MAXGLOBS);
   if (nParm>0) {
      if (GetConfInt(cParm,nParm,"TRACE",&i)) trace = i;
      if (GetConfBool(cParm,nParm,"SAVEBINARY",&b)) saveBinary = b;
   }
}

void ReportUsage(void)
{
   printf("\nUSAGE: HEAdapt [options] hmmList dataFiles...\n\n");
   printf(" Option                                       Default\n\n");
   printf(" -b N    set no. of blocks for transform      1\n");
   printf(" -c f    Mixture pruning threshold            10.0\n");
   printf(" -d s    dir to find hmm definitions          current\n");
   printf(" -f s1 s2  Save s2 for field s1 in tmf        defaults saved\n");
   printf(" -g      do global adaptation only            reg classes\n");
   printf(" -i N    update the transforms every N utts   static\n");
   printf(" -j f    Map adaptation with scaling factor   off\n");
   printf(" -k      Use MLLR before performing MAP       off\n");
   printf(" -m f    occ count min. to calc transform     700.0\n");
   printf(" -o s    extension for new hmm files          as src\n");
   printf(" -t f [i l] set pruning to f [inc limit]      inf\n");
   printf(" -u [mv] update means(m) vars(v)              means\n");
   printf(" -v f    set minimum variance to f            0.0\n");
   printf(" -x s    extension for hmm files              none\n");
   PrintStdOpts("BFGHIJKLMSTX");
   printf("\n\n");
}

void SetuFlags(void)
{
   char *s;
   Boolean meanFlag=FALSE;
   
   s=GetStrArg();
   while (*s != '\0')
      switch (*s++) {
      case 'm':
         meanFlag = TRUE;
         if (regTrans == TRANSUNDEF)
            regTrans=MEANONLY;
         break;
      case 'v':regTrans=MEANVAR;break;
      default: HError(2720,"SetuFlags: Unknown update flag %c",*s);
         break;
      }
   if (!meanFlag)
      HError(-2721,"SetuFlags: Mean transformation will be updated, even though unrequestred");
}

int main(int argc, char *argv[])
{
   int numUtt;
   char fmt[MAXSTRLEN];
   char *datafn, *s, *c;
   UttInfo *utt;            /* utterance information storage */
   RegTransInfo *rt;        /* regression transform information storage */
   HMMSet hset;             /* Set of HMMs to be re-estimated */
   FBInfo *fbInfo;          /* forward-backward information storage */

   void Initialise(FBInfo *fbInfo, MemHeap *x, HMMSet *hset,
                   char *hmmListFn, RegTransInfo *rt, char *transFile);
   void DoForwardBackward(FBInfo *fbInfo, UttInfo *utt, char *datafn, char *datafn2);
   
   if(InitShell(argc,argv,headapt_version,headapt_vc_id)<SUCCESS)
      HError(2700,"HEAdapt: InitShell failed");

   InitMem();
   InitMath();
   InitSigP();
   InitAudio();
   InitWave();
   InitVQ();
   InitModel();
   if(InitParm()<SUCCESS)  
      HError(2700,"HEAdapt: InitParm failed"); 
   InitLabel();
   InitTrain();
   InitUtil();
   InitFB();
   InitAdapt();  

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

   SetConfParms();
   CreateHeap(&hmmStack,"HmmStore", MSTAK, 1, 1.0, 50000, 500000);
   CreateHMMSet(&hset,&hmmStack,TRUE);
   CreateHeap(&uttStack,   "uttStore",    MSTAK, 1, 0.5, 100,   1000);
   utt = (UttInfo *) New(&uttStack, sizeof(UttInfo));
   CreateHeap(&fbInfoStack,   "FBInfoStore",  MSTAK, 1, 0.5, 1000,  10000);
   fbInfo = (FBInfo *) New(&fbInfoStack, sizeof(FBInfo));
   CreateHeap(&regStack,   "regClassStore",    MSTAK, 1, 0.5, 500,   5000);
   rt = (RegTransInfo *) New(&regStack, sizeof(RegTransInfo));

   while (NextArg() == SWITCHARG) {    
      s = GetSwtArg();
      if (strlen(s)!=1) 
         HError(2719,"HEAdapt: Bad switch %s; must be single letter",s);
      switch(s[0]){
      case 'b':
         if (NextArg()!= INTARG)
            HError(2719,"HEAdapt: Number of blocks in transform matrix expected");
         blocks = GetIntArg(); break; 
      case 'c':
         minFrwdP = GetChkedFlt(0.0,1000.0,s);
         break;
      case 'd':
         if (NextArg()!=STRINGARG)
            HError(2719,"HEAdapt: HMM definition directory expected");
         hmmDir = GetStrArg(); break;   
      case 'f':
         if (NextArg()!=STRINGARG)
            HError(2719,"HEAdapt: TMF description type expected");
         strcpy(fmt,GetStrArg());
         for (c=fmt; *c!=0; *c++=toupper(*c));
         if (strcmp(fmt,"UID")==0) {
            if (NextArg()!=STRINGARG)
               HError(2719,"HEAdapt: user id string expected");
            uid = GetStrArg();
         }
         else if (strcmp(fmt,"UNAME")==0) {
            if (NextArg()!=STRINGARG)
               HError(2719,"HEAdapt: user name string expected");
            uname = GetStrArg();
         }
         else if (strcmp(fmt,"CHAN")==0) {
            if (NextArg()!=STRINGARG)