main_decode_anytopo.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 1,466 行 · 第 1/4 页

		
		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, matchexact, ascr + lscr);
		
		printf ("BSTXCT: ");
		log_hypseg (uttid, stdout, hyp, nfr, scl, lwf);
	    }
	    
	    dag_destroy ();
	}
	
	lm_cache_stats_dump (lm);
	lm_cache_reset (lm);
    } else {
	E_ERROR ("%s: Viterbi search failed\n", uttid);
	hyp = NULL;
    }
    
    /* Log recognition output to the standard match and matchseg files */
    if (matchfp)
	log_hypstr (matchfp, hyp, uttid, matchexact, 0);
    if (matchsegfp)
	log_hypseg (uttid, matchsegfp, hyp, nfr, scl, lwf);

    ptmr_stop (&tmr_utt);
    
    printf ("%s: ", uttid);

    pctr_print(stderr,ctr_nfrm);
    pctr_print(stderr,ctr_nsen);

    printf ("%s: TMR: %5d Frm", uttid, nfr);
    if (nfr > 0) {
	printf (" %6.2f xEl", tmr_utt.t_elapsed * 100.0 / nfr);
	printf (" %6.2f xCPU", tmr_utt.t_cpu * 100.0 / nfr);

	if (tmr_utt.t_cpu > 0.0) {
	    printf (" [fwd %6.2fx %3d%%]", tmr_fwdvit.t_cpu * 100.0 / nfr,
		    (int32) ((tmr_fwdvit.t_cpu * 100.0) / tmr_utt.t_cpu));
	    printf ("[gau+sen %6.2fx %2d%%]", tmr_gausen.t_cpu * 100.0 / nfr,
		    (int32) ((tmr_gausen.t_cpu * 100.0) / tmr_utt.t_cpu));
	    printf ("[srch %6.2fx %2d%%]", tmr_fwdsrch.t_cpu * 100.0 / nfr,
		    (int32) ((tmr_fwdsrch.t_cpu * 100.0) / tmr_utt.t_cpu));
	    
	    fwd_timing_dump (tmr_utt.t_cpu);
	    
	    if (bp)
		printf ("[bp %6.2fx %2d%%]", tmr_bstpth.t_cpu * 100.0 / nfr,
			(int32) ((tmr_bstpth.t_cpu * 100.0) / tmr_utt.t_cpu));
	}
    }
    printf ("\n");
    fflush (stdout);
    
    tot_nfr += nfr;
}


/* Process utterances in the control file (-ctl argument) */
static int32 process_ctlfile ( void )
{
    FILE *ctlfp, *mllrctlfp;
    char *ctlfile, *cepdir, *cepext, *mllrctlfile;
    char *matchfile, *matchsegfile;
    char line[1024], ctlspec[1024], cepfile[1024], uttid[1024];
    char mllrfile[1024], prevmllr[1024];
/* CHANGE BY BHIKSHA: ADDED veclen AS A VARIABLE, 6 JAN 98 */
    int32 ctloffset, ctlcount, veclen, sf, ef, nfr;
/* END OF CHANGES BY BHIKSHA */
    int32 err_status;
    int32 i, k;
    
    err_status = 0;
    
    if ((ctlfile = (char *) cmd_ln_access("-ctl")) == NULL)
	E_FATAL("No -ctl argument\n");
    E_INFO("Processing ctl file %s\n", ctlfile);
    
    if ((mllrctlfile = (char *) cmd_ln_access("-mllrctl")) != NULL) {
	if ((mllrctlfp = fopen (mllrctlfile, "r")) == NULL)
	    E_FATAL("fopen(%s,r) failed\n", mllrctlfile);
    } else
	mllrctlfp = NULL;
    prevmllr[0] = '\0';
    
    if ((ctlfp = fopen (ctlfile, "r")) == NULL)
	E_FATAL("fopen(%s,r) failed\n", ctlfile);
    
    if ((matchfile = (char *) cmd_ln_access("-hyp")) == NULL) {
	matchfp = NULL;
    } else {
	/* Look for ,EXACT suffix, for retaining fillers/pronunciation specs in output */
	k = strlen (matchfile);
	if ((k > 6) && (strcmp (matchfile+(k-6), ",EXACT") == 0)) {
	    matchexact = 1;
	    matchfile[k-6] = '\0';
	} else
	    matchexact = 0;

	if ((matchfp = fopen (matchfile, "w")) == NULL)
	    E_ERROR("fopen(%s,w) failed\n", matchfile);
    }
    
    if ((matchsegfile = (char *) cmd_ln_access("-hypseg")) == NULL) {
	E_WARN("No -hypseg argument\n");
	matchsegfp = NULL;
    } else {
	if ((matchsegfp = fopen (matchsegfile, "w")) == NULL)
	    E_ERROR("fopen(%s,w) failed\n", matchsegfile);
    }
    
    cepdir = (char *) cmd_ln_access("-cepdir");
    cepext = (char *) cmd_ln_access("-cepext");
    assert ((cepdir != NULL) && (cepext != NULL));

/* BHIKSHA: ADDING VECLEN TO ALLOW VECTORS OF DIFFERENT SIZES */
    veclen = *((int32 *) cmd_ln_access("-ceplen"));
/* END CHANGES, 6 JAN 1998, BHIKSHA */
    
    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 err_status;
    }
    
    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) {
	    if (mllrctlfp) {
		if (fscanf (mllrctlfp, "%s", mllrfile) != 1)
		    E_FATAL ("Unexpected EOF(%s)\n", mllrctlfile);
	    }
	    --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);
	}

	if (mllrctlfp) {
	    if (fscanf (mllrctlfp, "%s", mllrfile) != 1)
		E_FATAL ("Unexpected EOF(%s)\n", mllrctlfile);
	    
	    if (strcmp (prevmllr, mllrfile) != 0) {
		float32 ***A, **B;
		int32 gid, sid;
		uint8 *mgau_xform;
		
		gauden_mean_reload (g, (char *) cmd_ln_access("-mean"));
		
		if (ms_mllr_read_regmat (mllrfile, &A, &B, fcb->stream_len,feat_n_stream(fcb)) < 0)
		    E_FATAL("ms_mllr_read_regmat failed\n");
		
		mgau_xform = (uint8 *) ckd_calloc (g->n_mgau, sizeof(uint8));

		/* Transform each non-CI mixture Gaussian */
		for (sid = 0; sid < sen->n_sen; sid++) {
		    if (mdef->cd2cisen[sid] != sid) {	/* Otherwise it's a CI senone */
			gid = sen->mgau[sid];
			if (! mgau_xform[gid]) {
			    ms_mllr_norm_mgau (g->mean[gid], g->n_density, A, B,
					    fcb->stream_len,feat_n_stream(fcb));
			    mgau_xform[gid] = 1;
			}
		    }
		}

		ckd_free (mgau_xform);
		
		ms_mllr_free_regmat (A, B, fcb->stream_len,feat_n_stream(fcb));

		strcpy (prevmllr, mllrfile);
	    }
	}

	if (! feat)
	  feat = feat_array_alloc (fcb, S3_MAX_FRAMES);
	
	/* Read and process mfc/feature speech input file */
	nfr = feat_s2mfc2feat(fcb, ctlspec, cepdir, cepext,sf, ef, feat, S3_MAX_FRAMES);
	assert(feat);

	if (nfr <= 0)
	    E_ERROR("Utt %s: Input file read (%s) failed\n", uttid, cepfile);
	else {

	  E_INFO ("%s: %d mfc frames\n", uttid, nfr);
	  assert(feat);
	  decode_utt (nfr, uttid);
	}
#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;
	}
    }
    
    if (matchfp)
	fclose (matchfp);
    if (matchsegfp)
	fclose (matchsegfp);

    fclose (ctlfp);
    if (mllrctlfp)
	fclose (mllrctlfp);
    
    return (err_status);
}


int main (int32 argc, char *argv[])
{
    int32 err_status;

    print_appl_info(argv[0]);
    cmd_ln_appl_enter(argc,argv,"default.arg",defn);
    unlimit ();
    
    if ((cmd_ln_access("-mdef") == NULL) ||
	(cmd_ln_access("-mean") == NULL) ||
	(cmd_ln_access("-var") == NULL)  ||
	(cmd_ln_access("-mixw") == NULL)  ||
	(cmd_ln_access("-tmat") == NULL))
	E_FATAL("Missing -mdef, -mean, -var, -mixw, or -tmat argument\n");

    if ((cmd_ln_access("-dict") == NULL) ||
	(cmd_ln_access("-lm") == NULL))
	E_FATAL("Missing -dict or -lm argument\n");
    
    inlatdir = (char *) cmd_ln_access ("-inlatdir");
    outlatdir = (char *) cmd_ln_access ("-outlatdir");
    if (outlatdir) {
	int32 k;
	
	k = strlen(outlatdir);
	if ((k > 6) && (strcmp (outlatdir+(k-6), ",NODES") == 0)) {
	    outlat_onlynodes = 1;
	    outlatdir[k-6] = '\0';
	} else
	    outlat_onlynodes = 0;
    }

    if (inlatdir && outlatdir && (strcmp (inlatdir, outlatdir) == 0))
	E_FATAL("Input and output lattice directories are the same\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.
     */
    {
	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 > 1.01)
	    E_WARN("Logbase %e perhaps too large??\n", logbase);
	logs3_init ((float64) logbase);
    }

    /* Initialize feature stream type */
    /*feat_init ((char *) cmd_ln_access ("-feat"));*/

    fcb = feat_init ( (char *) cmd_ln_access ("-feat"),
		      (char *) cmd_ln_access ("-cmn"),
		      (char *) cmd_ln_access ("-varnorm"),
		      (char *) cmd_ln_access ("-agc"));
    
    /* Read in input databases */
    models_init ();
    
    /* Senone scaling factor in each frame */
    senscale = (int32 *) ckd_calloc (S3_MAX_FRAMES, sizeof(int32));

    /* Best statescore in each frame */
    bestscr = (int32 *) ckd_calloc (S3_MAX_FRAMES, sizeof(int32));
    
    /* Allocate profiling timers and counters */
    ptmr_init(&tmr_utt);
    ptmr_init(&tmr_fwdvit);
    ptmr_init(&tmr_bstpth);
    ptmr_init(&tmr_gausen);
    ptmr_init(&tmr_fwdsrch);
    
    pctr_new(ctr_nfrm,"frm");
    pctr_new(ctr_nsen,"sen");

    /* Initialize forward Viterbi search module */
    fwd_init (mdef,tmat,dict,lm);
    printf ("\n");
    
    tot_nfr = 0;
    
    err_status = process_ctlfile ();

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

#if (! WIN32)
#if defined(_SUN4)
    system("ps -el | grep s3decode");
#else
    system ("ps aguxwww | grep s3decode");
#endif
#endif


    cmd_ln_appl_exit();

    exit(err_status);
    return 0 ;
}