main_dag.c

来自「CMU大名鼎鼎的SPHINX－3大词汇量连续语音识别系统」· C语言 代码 · 共 664 行 · 第 1/2 页

/*
 * Write exact hypothesis.  Format:
 *   <id> T <scr> A <ascr> L <lscr> {<sf> <wascr> <wlscr> <word>}... <ef>
 * where:
 *   scr = ascr + (lscr*lw+N*wip), where N = #words excluding <s>
 *   ascr = scaled acoustic score for entire utterance
 *   lscr = LM score (without lw or wip) for entire utterance
 *   sf = start frame for word
 *   wascr = scaled acoustic score for word
 *   wlscr = LM score (without lw or wip) for word
 *   ef = end frame for utterance.
 */
static void log_hypseg (char *uttid,
			FILE *fp,	/* Out: output file */
			srch_hyp_t *hypptr,	/* In: Hypothesis */
			int32 nfrm)	/* In: #frames in utterance */
{
    srch_hyp_t *h;
    int32 ascr, lscr, tscr;
    
    ascr = lscr = tscr = 0;
    for (h = hypptr; h; h = h->next) {
	ascr += h->ascr;
	if (dict_basewid(dict,h->wid) != startwid) {
	    lscr += lm_rawscore (lm,h->lscr, 1.0);
	} else {
	    assert (h->lscr == 0);
	}
	tscr += h->ascr + h->lscr;
    }

    fprintf (fp, "%s T %d A %d L %d", uttid, tscr, ascr, lscr);
    
    if (! hypptr)	/* HACK!! */
	fprintf (fp, " (null)\n");
    else {
	for (h = hypptr; h; h = h->next) {
	    lscr = (dict_basewid(dict,h->wid) != startwid) ? lm_rawscore (lm,h->lscr, 1.0) : 0;
	    fprintf (fp, " %d %d %d %s", h->sf, h->ascr, lscr, dict_wordstr (dict,h->wid));
	}
	fprintf (fp, " %d\n", nfrm);
    }
    
    fflush (fp);
}


/* Write hypothesis in old (pre-Nov95) NIST format */
static void log_hypstr (FILE *fp, srch_hyp_t *hypptr, char *uttid, int32 scr)
{
    srch_hyp_t *h;
    s3wid_t w;

    if (! hypptr)	/* HACK!! */
	fprintf (fp, "(null)");
    
    for (h = hypptr; h; h = h->next) {
	w = dict_basewid (dict,h->wid);
	if ((w != startwid) && (w != finishwid) && (! dict_filler_word (dict,w)))
	    fprintf (fp, "%s ", dict_wordstr(dict,w));
    }

    if (scr != 0)
	fprintf (fp, " (%s %d)\n", uttid, scr);
    else
	fprintf (fp, " (%s)\n", uttid);

    fflush (fp);
}


/* Log hypothesis in detail with word segmentations, acoustic and LM scores  */
static void log_hyp_detailed (FILE *fp, srch_hyp_t *hypptr, char *uttid, char *LBL, char *lbl)
{
    srch_hyp_t *h;
    int32 ascr_norm, lscr;

    ascr_norm = 0;
    lscr = 0;
    
    fprintf (fp, "%s:%s> %20s %5s %5s %11s %10s\n", LBL, uttid,
	     "WORD", "SFrm", "EFrm", "AScr(Norm)", "LMScore");
    
    for (h = hypptr; h; h = h->next) {
	fprintf (fp, "%s:%s> %20s %5d %5d %11d %10d\n", lbl, uttid,
		 h->word, h->sf, h->ef, h->ascr, h->lscr);

	ascr_norm += h->ascr;
	lscr += h->lscr;
    }

    fprintf (fp, "%s:%s> %20s %5s %5s %11d %10d\n", LBL, uttid,
	     "TOTAL", "", "", ascr_norm, lscr);
}


/* Decode the given mfc file and write result to matchfp and matchsegfp */
static void decode_utt (char *uttid, FILE *matchfp, FILE *matchsegfp)
{
    char dagfile[1024];
    srch_hyp_t *h, *hyp;
    char *latdir, *latext;
    int32 nfrm, ascr, lscr;


    ptmr_reset (&tm_utt);
    ptmr_start (&tm_utt);

    
    latdir = (char *) cmd_ln_access ("-inlatdir");
    latext = (char *) cmd_ln_access ("-latext");
    if (latdir)
	sprintf (dagfile, "%s/%s.%s", latdir, uttid, latext);
    else
	sprintf (dagfile, "%s.%s", uttid, latext);

    
    if ((nfrm = s3dag_dag_load (dagfile)) >= 0) {
	hyp = s3dag_dag_search (uttid);
	if(hyp!=NULL){
	  if ( *((int32 *) cmd_ln_access("-backtrace")) )
	    log_hyp_detailed (stdout, hyp, uttid, "BP", "bp");
	  
	  /* Total scaled acoustic score and LM score */
	  ascr = lscr = 0;
	  for (h = hyp; h; h = h->next) {
	    ascr += h->ascr;
	    lscr += h->lscr;
	  }
	  
	  printf ("BSTPTH: ");
	  log_hypstr (stdout, hyp, uttid, ascr+lscr);
	  
	  printf ("BSTXCT: ");
	  log_hypseg (uttid, stdout, hyp, nfrm);
	  
	  lm_cache_stats_dump (lm);
	  lm_cache_reset (lm);
	}else{
	  E_ERROR("DAG search (%s) failed\n", uttid);
	  hyp = NULL;
	}
    } else {
	E_ERROR("DAG search (%s) failed\n", uttid);
	hyp = NULL;
    }

    
    /* Log recognition output to the standard match and matchseg files */
    if (matchfp)
	log_hypstr (matchfp, hyp, uttid, 0);
    if (matchsegfp)
	log_hypseg (uttid, matchsegfp, hyp, nfrm);
    
    dag_destroy ();

    ptmr_stop (&tm_utt);
    
    printf ("%s: TMR: %5d Frm", uttid, nfrm);
    if (nfrm > 0) {
	printf (" %6.2f xEl", tm_utt.t_elapsed * 100.0 / nfrm);
	printf (" %6.2f xCPU", tm_utt.t_cpu * 100.0 / nfrm);
    }
    printf ("\n");
    fflush (stdout);

    tot_nfr += nfrm;
}


/* Process utterances in the control file (-ctl argument) */
static void process_ctlfile ( void )
{
    FILE *ctlfp, *matchfp, *matchsegfp;
    char *ctlfile;
    char *matchfile, *matchsegfile;
    char line[1024], ctlspec[1024], uttid[1024];
    int32 ctloffset, ctlcount;
    int32 i, k, sf, ef;
    
    if ((ctlfile = (char *) cmd_ln_access("-ctl")) == NULL)
	E_FATAL("No -ctl argument\n");
    
    E_INFO("Processing ctl file %s\n", ctlfile);
    
    if ((ctlfp = fopen (ctlfile, "r")) == NULL)
	E_FATAL("fopen(%s,r) failed\n", ctlfile);
    
    if ((matchfile = (char *) cmd_ln_access("-match")) == NULL) {
	E_WARN("No -match argument\n");
	matchfp = NULL;
    } else {
	if ((matchfp = fopen (matchfile, "w")) == NULL)
	    E_ERROR("fopen(%s,w) failed\n", matchfile);
    }
    
    if ((matchsegfile = (char *) cmd_ln_access("-matchseg")) == NULL) {
	E_WARN("No -matchseg argument\n");
	matchsegfp = NULL;
    } else {
	if ((matchsegfp = fopen (matchsegfile, "w")) == NULL)
	    E_ERROR("fopen(%s,w) failed\n", matchsegfile);
    }
    
    ctloffset = *((int32 *) cmd_ln_access("-ctloffset"));
    if (! cmd_ln_access("-ctlcount"))
	ctlcount = 0x7fffffff;	/* All entries processed if no count specified */
    else
	ctlcount = *((int32 *) cmd_ln_access("-ctlcount"));
    if (ctlcount == 0) {
	E_INFO("-ctlcount argument = 0!!\n");
	fclose (ctlfp);
	return;
    }

    if (ctloffset > 0)
	E_INFO("Skipping %d utterances in the beginning of control file\n",
	       ctloffset);
    while ((ctloffset > 0) && (fgets(line, sizeof(line), ctlfp) != NULL)) {
	if (sscanf (line, "%s", ctlspec) > 0)
	    --ctloffset;
    }
    
    while ((ctlcount > 0) && (fgets(line, sizeof(line), ctlfp) != NULL)) {
	printf ("\n");
	E_INFO("Utterance: %s", line);

	sf = 0;
	ef = (int32)0x7ffffff0;
	if ((k = sscanf (line, "%s %d %d %s", ctlspec, &sf, &ef, uttid)) <= 0)
	    continue;	    /* Empty line */

	if ((k == 2) || ( (k >= 3) && ((sf >= ef) || (sf < 0))) ) {
	    E_ERROR("Error in ctlfile spec; skipped\n");
	    /* What happens to ctlcount??? */
	    continue;
	}
	if (k < 4) {
	    /* Create utt-id from mfc-filename (and sf/ef if specified) */
	    for (i = strlen(ctlspec)-1; (i >= 0) && (ctlspec[i] != '/'); --i);
	    if (k == 3)
		sprintf (uttid, "%s_%d_%d", ctlspec+i+1, sf, ef);
	    else
		strcpy (uttid, ctlspec+i+1);
	}


	decode_utt (uttid, matchfp, matchsegfp);

	--ctlcount;
    }
    printf ("\n");

    if (fscanf (ctlfp, "%s", line) == 1)
	E_INFO("Skipping rest of control file beginning with:\n\t%s\n", line);

    if (matchfp)
	fclose (matchfp);
    if (matchsegfp)
	fclose (matchsegfp);

    fclose (ctlfp);
}

int main (int32 argc, char *argv[])
{
  /*  kb_t kb;
      ptmr_t tm;*/

  print_appl_info(argv[0]);
  cmd_ln_appl_enter(argc,argv,"default.arg",defn);

  unlimit ();
    
  if ((cmd_ln_access("-mdef") == NULL) ||
      (cmd_ln_access("-dict") == NULL) ||
      (cmd_ln_access("-lm") == NULL))
    E_FATAL("Missing -mdef, -dict, or -lm argument\n");
  
  /*
   * Initialize log(S3-base).  All scores (probs...) computed in log domain to avoid
   * underflow.  At the same time, log base = 1.0001 (1+epsilon) to allow log values
   * to be maintained in int32 variables without significant loss of precision.
   */
  if (cmd_ln_access("-logbase") == NULL)
    logs3_init (1.0001);
  else {
    float32 logbase;
    
    logbase = *((float32 *) cmd_ln_access("-logbase"));
    if (logbase <= 1.0)
      E_FATAL("Illegal log-base: %e; must be > 1.0\n", logbase);
    if (logbase > 1.1)
      E_WARN("Logbase %e perhaps too large??\n", logbase);
    logs3_init ((float64) logbase);
  }
    
  /* Read in input databases */
  models_init ();
  
  /* Allocate timing object */
  ptmr_init(&tm_utt);
  tot_nfr = 0;
    
  /* Initialize forward Viterbi search module */
  dag_init (dict);
  printf ("\n");
  
  process_ctlfile ();

  printf ("\n");
  printf("TOTAL FRAMES:       %8d\n", tot_nfr);
  if (tot_nfr > 0) {
    printf("TOTAL CPU TIME:     %11.2f sec, %7.2f xRT\n",
	   tm_utt.t_tot_cpu, tm_utt.t_tot_cpu/(tot_nfr*0.01));
    printf("TOTAL ELAPSED TIME: %11.2f sec, %7.2f xRT\n",
	   tm_utt.t_tot_elapsed, tm_utt.t_tot_elapsed/(tot_nfr*0.01));
  }
  fflush (stdout);

#if (! WIN32)
    system ("ps auxwww | grep s3dag");
#endif

    /* Hack!! To avoid hanging problem under Linux */

    cmd_ln_appl_exit();
    
    return 0;
}