main_astar.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 497 行 · 第 1/2 页

      "Input word-lattice directory with per-utt files for restricting words searched" },
    { "-latext",
      ARG_STRING,
      "lat.gz",
      "Word-lattice filename extension (.gz or .Z extension for compression)" },
    { "-nbestdir",
      ARG_STRING,
      ".",
      "Input word-lattice directory with per-utt files for restricting words searched" },
    { "-nbestext",
      ARG_STRING,
      "nbest.gz",
      "N-best filename extension (.gz or .Z extension for compression)" },
    { "-nbest",
      ARG_INT32,
      "200",
      "Max. n-best hypotheses to generate per utterance" },
    { "-min_endfr",
      ARG_INT32,
      "3",
      "Nodes ignored during search if they persist for fewer than so many end frames" },
    { "-dagfudge",
      ARG_INT32,
      "2",
      "Adjacency fudge (#frames) between nodes in DAG (0..2)" },
    { "-ppathdebug",
      ARG_INT32,
      "0",
      "Adjacency fudge (#frames) between nodes in DAG (0..2)" },
    { "-logfn",
      ARG_STRING,
      NULL,
      "Log file (default stdout/stderr)" },
    
    { NULL, ARG_INT32, NULL, NULL }
};

/*
 * Load and cross-check all models (acoustic/lexical/linguistic).
 */
static void models_init ( void )
{
    /* HMM model definition */
    mdef = mdef_init ((char *) cmd_ln_access("-mdef"));

    /* Dictionary */
    dict = dict_init (mdef, (char *) cmd_ln_access("-dict"),
		      (char *) cmd_ln_access("-fdict"), 0);

    /* HACK!! Make sure SILENCE_WORD, START_WORD and FINISH_WORD are in dictionary */
    silwid = dict_wordid (dict, S3_SILENCE_WORD);
    startwid = dict_wordid (dict, S3_START_WORD);
    finishwid = dict_wordid (dict, S3_FINISH_WORD);
    if (NOT_S3WID(silwid) || NOT_S3WID(startwid) || NOT_S3WID(finishwid)) {
	E_FATAL("%s, %s, or %s missing from dictionary\n",
		S3_SILENCE_WORD, S3_START_WORD, S3_FINISH_WORD);
    }
    if ((dict->filler_start > dict->filler_end) || (! dict_filler_word (dict, silwid)))
	E_FATAL("%s must occur (only) in filler dictionary\n", S3_SILENCE_WORD);
    /* No check that alternative pronunciations for filler words are in filler range!! */

    /* LM */
    {
	char *lmfile;
	
	lmfile = (char *) cmd_ln_access("-lm");
	if (! lmfile)
	    E_FATAL("-lm argument missing\n");
	lm = lm_read (lmfile, *(float32 *)cmd_ln_access("-lw"),
	    		      *(float32 *)cmd_ln_access("-inspen"),
			      *(float32 *)cmd_ln_access("-ugwt"));

	fpen = fillpen_init (dict, (char *) cmd_ln_access("-fillpen"),
		      *(float32 *)cmd_ln_access("-silpen"),
		      *(float32 *)cmd_ln_access("-noisepen"),
		      *(float32 *)cmd_ln_access("-lw"),
		      *(float32 *)cmd_ln_access("-inspen"));
    }

    dict2lmwid = wid_dict_lm_map(dict, lm, *(float32 *)cmd_ln_access("-lw"));
}


/* Decode the given mfc file and write result to given directory */
static void decode_utt (char *uttid, char *nbestdir)
{
    char dagfile[1024], nbestfile[1024];
    char *latdir, *latext, *nbestext;
    int32 nfrm;
    
    latdir = (char *) cmd_ln_access ("-inlatdir");
    latext = (char *) cmd_ln_access ("-latext");
    nbestext = (char *) cmd_ln_access ("-nbestext");
    if (latdir)
	sprintf (dagfile, "%s/%s.%s", latdir, uttid, latext);
    else
	sprintf (dagfile, "%s.%s", uttid, latext);
    
    ptmr_reset (&tm_utt);
    ptmr_start (&tm_utt);

    if ((nfrm = s3astar_dag_load (dagfile)) > 0) {
	sprintf (nbestfile, "%s/%s.%s", nbestdir, uttid, nbestext);
	nbest_search (nbestfile, uttid);
	
	lm_cache_stats_dump (lm);
	lm_cache_reset (lm);
    } else
	E_ERROR("Dag load (%s) failed\n", uttid);
    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);
}


/* Process utterances in the control file (-ctl argument) */
static void process_ctlfile ( void )
{
    FILE *ctlfp;
    char *ctlfile, *nbestdir;
    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);
    
    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;
    }
    
    nbestdir = (char *) cmd_ln_access ("-nbestdir");

    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, nbestdir);
#if 0
	linklist_stats ();
#endif
	--ctlcount;
    }
    printf ("\n");

    while (fgets(line, sizeof(line), ctlfp) != NULL) {
	if (sscanf (line, "%s", ctlspec) > 0) {
	    E_INFO("Skipping rest of control file beginning with:\n\t%s", line);
	    break;
	}
    }

    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 ();
    
    /*
     * 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.
     */
    {
	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 ();

    /* Initialize forward Viterbi search module */
    nbest_init ();
    printf ("\n");

    ptmr_init (&tm_utt);
    
    process_ctlfile ();

#if (! WIN32)
    system ("ps aguxwww | grep s3astar");
#endif

    cmd_ln_appl_exit();
    return 0;
}