1 /* -*- c-basic-offset: 4; indent-tabs-mode: nil -*- */
2 /* ====================================================================
3 * Copyright (c) 1999-2004 Carnegie Mellon University. All rights
6 * Redistribution and use in source and binary forms, with or without
7 * modification, are permitted provided that the following conditions
10 * 1. Redistributions of source code must retain the above copyright
11 * notice, this list of conditions and the following disclaimer.
13 * 2. Redistributions in binary form must reproduce the above copyright
14 * notice, this list of conditions and the following disclaimer in
15 * the documentation and/or other materials provided with the
18 * This work was supported in part by funding from the Defense Advanced
19 * Research Projects Agency and the National Science Foundation of the
20 * United States of America, and the CMU Sphinx Speech Consortium.
22 * THIS SOFTWARE IS PROVIDED BY CARNEGIE MELLON UNIVERSITY ``AS IS'' AND
23 * ANY EXPRESSED OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO,
24 * THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
25 * PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL CARNEGIE MELLON UNIVERSITY
26 * NOR ITS EMPLOYEES BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
27 * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
28 * LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
29 * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
30 * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
31 * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
32 * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
34 * ====================================================================
39 * mdef.c -- HMM model definition: base (CI) phones and triphones
41 * **********************************************
42 * CMU ARPA Speech Project
44 * Copyright (c) 1999 Carnegie Mellon University.
45 * ALL RIGHTS RESERVED.
46 * **********************************************
51 * 22-Nov-2004 M K Ravishankar (rkm@cs.cmu.edu) at Carnegie Mellon University
52 * Imported from s3.2, for supporting s3 format continuous
55 * 14-Oct-1999 M K Ravishankar (rkm@cs.cmu.edu) at Carnegie Mellon
56 * Added mdef_sseq2sen_active().
58 * 06-May-1999 M K Ravishankar (rkm@cs.cmu.edu) at Carnegie Mellon
59 * In mdef_phone_id(), added backing off to silence phone context from filler
60 * context if original triphone not found.
62 * 30-Apr-1999 M K Ravishankar (rkm@cs.cmu.edu) at Carnegie Mellon
63 * Added senone-sequence id (ssid) to phone_t and appropriate functions to
64 * maintain it. Instead, moved state sequence info to mdef_t.
66 * 13-Jul-96 M K Ravishankar (rkm@cs.cmu.edu) at Carnegie Mellon University.
67 * Added mdef_phone_str().
69 * 01-Jan-96 M K Ravishankar (rkm@cs.cmu.edu) at Carnegie Mellon University.
70 * Allowed mdef_phone_id_nearest to return base phone id if either
71 * left or right context (or both) is undefined.
73 * 01-Jan-96 M K Ravishankar (rkm@cs.cmu.edu) at Carnegie Mellon University.
80 * All phones have same #states, same topology.
81 * Every phone has exactly one non-emitting, final state--the last one.
82 * CI phones must appear first in model definition file.
91 /* SphinxBase headers. */
92 #include <sphinxbase/ckd_alloc.h>
93 #include <sphinxbase/err.h>
99 #define MODEL_DEF_VERSION "0.3"
102 ciphone_add(mdef_t * m, char *ci, int p)
104 assert(p < m->n_ciphone);
106 m->ciphone[p].name = (char *) ckd_salloc(ci); /* freed in mdef_free */
107 if (hash_table_enter(m->ciphone_ht, m->ciphone[p].name,
108 (void *)(long)p) != (void *)(long)p)
109 E_FATAL("hash_table_enter(%s) failed; duplicate CIphone?\n",
115 find_ph_lc(ph_lc_t * lclist, int lc)
119 for (lcptr = lclist; lcptr && (lcptr->lc != lc); lcptr = lcptr->next);
125 find_ph_rc(ph_rc_t * rclist, int rc)
129 for (rcptr = rclist; rcptr && (rcptr->rc != rc); rcptr = rcptr->next);
135 triphone_add(mdef_t * m,
136 int ci, int lc, int rc, word_posn_t wpos,
142 assert(p < m->n_phone);
144 /* Fill in phone[p] information (state and tmat mappings added later) */
148 m->phone[p].wpos = wpos;
150 /* Create <ci,lc,rc,wpos> -> p mapping if not a CI phone */
151 if (p >= m->n_ciphone) {
152 if ((lcptr = find_ph_lc(m->wpos_ci_lclist[wpos][(int) ci], lc))
154 lcptr = (ph_lc_t *) ckd_calloc(1, sizeof(ph_lc_t)); /* freed at mdef_free, I believe */
156 lcptr->next = m->wpos_ci_lclist[wpos][(int) ci];
157 m->wpos_ci_lclist[wpos][(int) ci] = lcptr; /* This is what needs to be freed */
159 if ((rcptr = find_ph_rc(lcptr->rclist, rc)) != NULL) {
160 __BIGSTACKVARIABLE__ char buf[4096];
162 mdef_phone_str(m, rcptr->pid, buf);
163 E_FATAL("Duplicate triphone: %s\n", buf);
166 rcptr = (ph_rc_t *) ckd_calloc(1, sizeof(ph_rc_t)); /* freed in mdef_free, I believe */
169 rcptr->next = lcptr->rclist;
170 lcptr->rclist = rcptr;
176 mdef_ciphone_id(mdef_t * m, char *ci)
179 if (hash_table_lookup_int32(m->ciphone_ht, ci, &id) < 0)
186 mdef_ciphone_str(mdef_t * m, int id)
189 assert((id >= 0) && (id < m->n_ciphone));
191 return (m->ciphone[id].name);
196 mdef_phone_str(mdef_t * m, int pid, char *buf)
201 assert((pid >= 0) && (pid < m->n_phone));
202 wpos_name = WPOS_NAME;
205 if (pid < m->n_ciphone)
206 sprintf(buf, "%s", mdef_ciphone_str(m, pid));
208 sprintf(buf, "%s %s %s %c",
209 mdef_ciphone_str(m, m->phone[pid].ci),
210 mdef_ciphone_str(m, m->phone[pid].lc),
211 mdef_ciphone_str(m, m->phone[pid].rc),
212 wpos_name[m->phone[pid].wpos]);
219 mdef_phone_id(mdef_t * m,
220 int ci, int lc, int rc, word_posn_t wpos)
227 assert((ci >= 0) && (ci < m->n_ciphone));
228 assert((lc >= 0) && (lc < m->n_ciphone));
229 assert((rc >= 0) && (rc < m->n_ciphone));
230 assert((wpos >= 0) && (wpos < N_WORD_POSN));
233 find_ph_lc(m->wpos_ci_lclist[wpos][(int) ci], lc)) == NULL)
234 || ((rcptr = find_ph_rc(lcptr->rclist, rc)) == NULL)) {
235 /* Not found; backoff to silence context if non-silence filler context */
239 newl = m->ciphone[(int) lc].filler ? m->sil : lc;
240 newr = m->ciphone[(int) rc].filler ? m->sil : rc;
241 if ((newl == lc) && (newr == rc))
244 return (mdef_phone_id(m, ci, newl, newr, wpos));
251 mdef_is_ciphone(mdef_t * m, int p)
254 assert((p >= 0) && (p < m->n_phone));
256 return ((p < m->n_ciphone) ? 1 : 0);
260 mdef_is_cisenone(mdef_t * m, int s)
267 return ((s == m->cd2cisen[s]) ? 1 : 0);
271 /* Parse tmat and state->senone mappings for phone p and fill in structure */
273 parse_tmat_senmap(mdef_t * m, char *line, int32 off, int p)
276 __BIGSTACKVARIABLE__ char word[1024], *lp;
280 /* Read transition matrix id */
281 if ((sscanf(lp, "%d%n", &n, &wlen) != 1) || (n < 0))
282 E_FATAL("Missing or bad transition matrix id: %s\n", line);
283 m->phone[p].tmat = n;
285 E_FATAL("tmat-id(%d) > #tmat in header(%d): %s\n", n, m->n_tmat,
289 /* Read senone mappings for each emitting state */
290 for (n = 0; n < m->n_emit_state; n++) {
291 if ((sscanf(lp, "%d%n", &s, &wlen) != 1) || (s < 0))
292 E_FATAL("Missing or bad state[%d]->senone mapping: %s\n", n,
295 /*20040821 ARCHAN, This line is added to allow 3.x/3.0 compatability. */
296 m->phone[p].state[n] = s;
298 if ((p < m->n_ciphone) && (m->n_ci_sen <= s))
299 E_FATAL("CI-senone-id(%d) > #CI-senones(%d): %s\n", s,
302 E_FATAL("Senone-id(%d) > #senones(%d): %s\n", s, m->n_sen,
309 /* Check for the last non-emitting state N */
310 if ((sscanf(lp, "%s%n", word, &wlen) != 1) || (strcmp(word, "N") != 0))
311 E_FATAL("Missing non-emitting state spec: %s\n", line);
314 /* Check for end of line */
315 if (sscanf(lp, "%s%n", word, &wlen) == 1)
316 E_FATAL("Non-empty beyond non-emitting final state: %s\n", line);
321 parse_base_line(mdef_t * m, char *line, int p)
324 __BIGSTACKVARIABLE__ char word[1024], *lp;
329 /* Read base phone name */
330 if (sscanf(lp, "%s%n", word, &wlen) != 1)
331 E_FATAL("Missing base phone name: %s\n", line);
334 /* Make sure it's not a duplicate */
335 ci = mdef_ciphone_id(m, word);
337 E_FATAL("Duplicate base phone: %s\n", line);
339 /* Add ciphone to ciphone table with id p */
340 ciphone_add(m, word, p);
343 /* Read and skip "-" for lc, rc, wpos */
344 for (n = 0; n < 3; n++) {
345 if ((sscanf(lp, "%s%n", word, &wlen) != 1)
346 || (strcmp(word, "-") != 0))
347 E_FATAL("Bad context info for base phone: %s\n", line);
351 /* Read filler attribute, if present */
352 if (sscanf(lp, "%s%n", word, &wlen) != 1)
353 E_FATAL("Missing filler atribute field: %s\n", line);
355 if (strcmp(word, "filler") == 0)
356 m->ciphone[(int) ci].filler = 1;
357 else if (strcmp(word, "n/a") == 0)
358 m->ciphone[(int) ci].filler = 0;
360 E_FATAL("Bad filler attribute field: %s\n", line);
362 triphone_add(m, ci, -1, -1, WORD_POSN_UNDEFINED, p);
364 /* Parse remainder of line: transition matrix and state->senone mappings */
365 parse_tmat_senmap(m, line, lp - line, p);
370 parse_tri_line(mdef_t * m, char *line, int p)
373 __BIGSTACKVARIABLE__ char word[1024], *lp;
375 word_posn_t wpos = WORD_POSN_BEGIN;
379 /* Read base phone name */
380 if (sscanf(lp, "%s%n", word, &wlen) != 1)
381 E_FATAL("Missing base phone name: %s\n", line);
384 ci = mdef_ciphone_id(m, word);
386 E_FATAL("Unknown base phone: %s\n", line);
389 if (sscanf(lp, "%s%n", word, &wlen) != 1)
390 E_FATAL("Missing left context: %s\n", line);
392 lc = mdef_ciphone_id(m, word);
394 E_FATAL("Unknown left context: %s\n", line);
397 if (sscanf(lp, "%s%n", word, &wlen) != 1)
398 E_FATAL("Missing right context: %s\n", line);
400 rc = mdef_ciphone_id(m, word);
402 E_FATAL("Unknown right context: %s\n", line);
404 /* Read tripone word-position within word */
405 if ((sscanf(lp, "%s%n", word, &wlen) != 1) || (word[1] != '\0'))
406 E_FATAL("Missing or bad word-position spec: %s\n", line);
410 wpos = WORD_POSN_BEGIN;
413 wpos = WORD_POSN_END;
416 wpos = WORD_POSN_SINGLE;
419 wpos = WORD_POSN_INTERNAL;
422 E_FATAL("Bad word-position spec: %s\n", line);
425 /* Read filler attribute, if present. Must match base phone attribute */
426 if (sscanf(lp, "%s%n", word, &wlen) != 1)
427 E_FATAL("Missing filler attribute field: %s\n", line);
429 if (((strcmp(word, "filler") == 0) && (m->ciphone[(int) ci].filler)) ||
430 ((strcmp(word, "n/a") == 0) && (!m->ciphone[(int) ci].filler))) {
431 /* Everything is fine */
434 E_FATAL("Bad filler attribute field: %s\n", line);
436 triphone_add(m, ci, lc, rc, wpos, p);
438 /* Parse remainder of line: transition matrix and state->senone mappings */
439 parse_tmat_senmap(m, line, lp - line, p);
444 sseq_compress(mdef_t * m)
454 k = m->n_emit_state * sizeof(int16);
456 h = hash_table_new(m->n_phone, HASH_CASE_YES);
459 /* Identify unique senone-sequence IDs. BUG: tmat-id not being considered!! */
460 for (p = 0; p < m->n_phone; p++) {
461 /* Add senone sequence to hash table */
463 == (j = hash_table_enter_bkey_int32(h, (char *)m->sseq[p], k, n_sseq)))
466 m->phone[p].ssid = j;
469 /* Generate compacted sseq table */
470 sseq = ckd_calloc_2d(n_sseq, m->n_emit_state, sizeof(**sseq)); /* freed in mdef_free() */
472 g = hash_table_tolist(h, &j);
475 for (gn = g; gn; gn = gnode_next(gn)) {
476 he = (hash_entry_t *) gnode_ptr(gn);
477 j = (long)hash_entry_val(he);
478 memcpy(sseq[j], hash_entry_key(he), k);
482 /* Free the old, temporary senone sequence table, replace with compacted one */
483 ckd_free_2d(m->sseq);
492 noncomment_line(char *line, int32 size, FILE * fp)
494 while (fgets(line, size, fp) != NULL) {
503 * Initialize phones (ci and triphones) and state->senone mappings from .mdef file.
506 mdef_init(char *mdeffile, int32 breport)
509 int32 n_ci, n_tri, n_map, n;
510 __BIGSTACKVARIABLE__ char tag[1024], buf[1024];
517 E_FATAL("No mdef-file\n");
520 E_INFO("Reading model definition: %s\n", mdeffile);
522 m = (mdef_t *) ckd_calloc(1, sizeof(mdef_t)); /* freed in mdef_free */
524 if ((fp = fopen(mdeffile, "r")) == NULL)
525 E_FATAL_SYSTEM("Failed to open mdef file '%s' for reading", mdeffile);
527 if (noncomment_line(buf, sizeof(buf), fp) < 0)
528 E_FATAL("Empty file: %s\n", mdeffile);
530 if (strncmp(buf, "BMDF", 4) == 0 || strncmp(buf, "FDMB", 4) == 0) {
532 ("Found byte-order mark %.4s, assuming this is a binary mdef file\n",
538 if (strncmp(buf, MODEL_DEF_VERSION, strlen(MODEL_DEF_VERSION)) != 0)
539 E_FATAL("Version error: Expecing %s, but read %s\n",
540 MODEL_DEF_VERSION, buf);
542 /* Read #base phones, #triphones, #senone mappings defined in header */
550 if (noncomment_line(buf, sizeof(buf), fp) < 0)
551 E_FATAL("Incomplete header\n");
553 if ((sscanf(buf, "%d %s", &n, tag) != 2) || (n < 0))
554 E_FATAL("Error in header: %s\n", buf);
556 if (strcmp(tag, "n_base") == 0)
558 else if (strcmp(tag, "n_tri") == 0)
560 else if (strcmp(tag, "n_state_map") == 0)
562 else if (strcmp(tag, "n_tied_ci_state") == 0)
564 else if (strcmp(tag, "n_tied_state") == 0)
566 else if (strcmp(tag, "n_tied_tmat") == 0)
569 E_FATAL("Unknown header line: %s\n", buf);
570 } while ((n_ci < 0) || (n_tri < 0) || (n_map < 0) ||
571 (m->n_ci_sen < 0) || (m->n_sen < 0) || (m->n_tmat < 0));
573 if ((n_ci == 0) || (m->n_ci_sen == 0) || (m->n_tmat == 0)
574 || (m->n_ci_sen > m->n_sen))
575 E_FATAL("%s: Error in header\n", mdeffile);
577 /* Check typesize limits */
578 if (n_ci >= MAX_INT16)
579 E_FATAL("%s: #CI phones (%d) exceeds limit (%d)\n", mdeffile, n_ci,
581 if (n_ci + n_tri >= MAX_INT32) /* Comparison is always false... */
582 E_FATAL("%s: #Phones (%d) exceeds limit (%d)\n", mdeffile,
583 n_ci + n_tri, MAX_INT32);
584 if (m->n_sen >= MAX_INT16)
585 E_FATAL("%s: #senones (%d) exceeds limit (%d)\n", mdeffile,
586 m->n_sen, MAX_INT16);
587 if (m->n_tmat >= MAX_INT32) /* Comparison is always false... */
588 E_FATAL("%s: #tmats (%d) exceeds limit (%d)\n", mdeffile,
589 m->n_tmat, MAX_INT32);
591 m->n_emit_state = (n_map / (n_ci + n_tri)) - 1;
592 if ((m->n_emit_state + 1) * (n_ci + n_tri) != n_map)
594 ("Header error: n_state_map not a multiple of n_ci*n_tri\n");
596 /* Initialize ciphone info */
598 m->ciphone_ht = hash_table_new(n_ci, HASH_CASE_YES); /* With case-insensitive string names *//* freed in mdef_free */
599 m->ciphone = (ciphone_t *) ckd_calloc(n_ci, sizeof(ciphone_t)); /* freed in mdef_free */
601 /* Initialize phones info (ciphones + triphones) */
602 m->n_phone = n_ci + n_tri;
603 m->phone = (phone_t *) ckd_calloc(m->n_phone, sizeof(phone_t)); /* freed in mdef_free */
605 /* Allocate space for state->senone map for each phone */
606 senmap = ckd_calloc_2d(m->n_phone, m->n_emit_state, sizeof(**senmap)); /* freed in mdef_free */
607 m->sseq = senmap; /* TEMPORARY; until it is compressed into just the unique ones */
610 /**CODE DUPLICATION!*****************************************************************************************************/
611 /* Flat decoder-specific */
612 /* Allocate space for state->senone map for each phone */
614 /* ARCHAN 20040820, this sacrifice readability and may cause pointer
615 problems in future. However, this is a less evil than
616 duplication of code. This is trick point all the state mapping
617 to the global mapping and avoid duplicated memory.
620 /* S3 xwdpid_compress will compress the below list phone list.
623 /* ARCHAN, this part should not be used when one of the recognizer is used. */
625 (int16 *) ckd_calloc(m->n_phone * m->n_emit_state,
626 sizeof(*m->st2senmap));
627 for (p = 0; p < m->n_phone; p++)
628 m->phone[p].state = m->st2senmap + (p * m->n_emit_state);
629 /******************************************************************************************************/
632 /* Allocate initial space for <ci,lc,rc,wpos> -> pid mapping */
633 m->wpos_ci_lclist = (ph_lc_t ***) ckd_calloc_2d(N_WORD_POSN, m->n_ciphone, sizeof(ph_lc_t *)); /* freed in mdef_free */
636 * Read base phones and triphones. They'll simply be assigned a running sequence
637 * number as their "phone-id". If the phone-id < n_ci, it's a ciphone.
640 /* Read base phones */
641 for (p = 0; p < n_ci; p++) {
642 if (noncomment_line(buf, sizeof(buf), fp) < 0)
643 E_FATAL("Premature EOF reading CIphone %d\n", p);
644 parse_base_line(m, buf, p);
646 m->sil = mdef_ciphone_id(m, S3_SILENCE_CIPHONE);
648 /* Read triphones, if any */
649 for (; p < m->n_phone; p++) {
650 if (noncomment_line(buf, sizeof(buf), fp) < 0)
651 E_FATAL("Premature EOF reading phone %d\n", p);
652 parse_tri_line(m, buf, p);
655 if (noncomment_line(buf, sizeof(buf), fp) >= 0)
656 E_ERROR("Non-empty file beyond expected #phones (%d)\n",
659 /* Build CD senones to CI senones map */
660 if (m->n_ciphone * m->n_emit_state != m->n_ci_sen)
662 ("#CI-senones(%d) != #CI-phone(%d) x #emitting-states(%d)\n",
663 m->n_ci_sen, m->n_ciphone, m->n_emit_state);
664 m->cd2cisen = (int16 *) ckd_calloc(m->n_sen, sizeof(*m->cd2cisen)); /* freed in mdef_free */
666 m->sen2cimap = (int16 *) ckd_calloc(m->n_sen, sizeof(*m->sen2cimap)); /* freed in mdef_free */
668 for (s = 0; s < m->n_sen; s++)
669 m->sen2cimap[s] = -1;
670 for (s = 0; s < m->n_ci_sen; s++) { /* CI senones */
672 m->sen2cimap[s] = s / m->n_emit_state;
674 for (p = n_ci; p < m->n_phone; p++) { /* CD senones */
675 for (s = 0; s < m->n_emit_state; s++) {
677 ci = m->sseq[m->phone[p].ci][s];
678 m->cd2cisen[cd] = ci;
679 m->sen2cimap[cd] = m->phone[p].ci;
690 mdef_report(mdef_t * m)
692 E_INFO_NOFN("Initialization of mdef_t, report:\n");
694 ("%d CI-phone, %d CD-phone, %d emitstate/phone, %d CI-sen, %d Sen, %d Sen-Seq\n",
695 m->n_ciphone, m->n_phone - m->n_ciphone, m->n_emit_state,
696 m->n_ci_sen, m->n_sen, m->n_sseq);
701 /* RAH 4.23.01, Need to step down the ->next list to see if there are
702 any more things to free
707 /* RAH 4.19.01, Attempt to free memory that was allocated within this module
708 I have not verified that all the memory has been freed. I've taken only a
709 reasonable effort for now.
710 RAH 4.24.01 - verified that all memory is released.
713 mdef_free_recursive_lc(ph_lc_t * lc)
719 mdef_free_recursive_rc(lc->rclist);
722 mdef_free_recursive_lc(lc->next);
724 ckd_free((void *) lc);
728 mdef_free_recursive_rc(ph_rc_t * rc)
734 mdef_free_recursive_rc(rc->next);
736 ckd_free((void *) rc);
740 /* RAH, Free memory that was allocated in mdef_init
741 Rational purify shows that no leaks exist
745 mdef_free(mdef_t * m)
751 ckd_free((void *) m->sen2cimap);
753 ckd_free((void *) m->cd2cisen);
755 /* RAH, go down the ->next list and delete all the pieces */
756 for (i = 0; i < N_WORD_POSN; i++)
757 for (j = 0; j < m->n_ciphone; j++)
758 if (m->wpos_ci_lclist[i][j]) {
759 mdef_free_recursive_lc(m->wpos_ci_lclist[i][j]->next);
760 mdef_free_recursive_rc(m->wpos_ci_lclist[i][j]->
764 for (i = 0; i < N_WORD_POSN; i++)
765 for (j = 0; j < m->n_ciphone; j++)
766 if (m->wpos_ci_lclist[i][j])
767 ckd_free((void *) m->wpos_ci_lclist[i][j]);
770 if (m->wpos_ci_lclist)
771 ckd_free_2d((void *) m->wpos_ci_lclist);
773 ckd_free_2d((void *) m->sseq);
774 /* Free phone context */
776 ckd_free((void *) m->phone);
778 hash_table_free(m->ciphone_ht);
780 for (i = 0; i < m->n_ciphone; i++) {
781 if (m->ciphone[i].name)
782 ckd_free((void *) m->ciphone[i].name);
787 ckd_free((void *) m->ciphone);
790 ckd_free((void *) m->st2senmap);
792 ckd_free((void *) m);