PocketSphinx  0.6
ms_mgau.c
1 /* -*- c-basic-offset: 4; indent-tabs-mode: nil -*- */
2 /* ====================================================================
3  * Copyright (c) 1999-2004 Carnegie Mellon University. All rights
4  * reserved.
5  *
6  * Redistribution and use in source and binary forms, with or without
7  * modification, are permitted provided that the following conditions
8  * are met:
9  *
10  * 1. Redistributions of source code must retain the above copyright
11  * notice, this list of conditions and the following disclaimer.
12  *
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
16  * distribution.
17  *
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.
21  *
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.
33  *
34  * ====================================================================
35  *
36  */
37 /*
38  * ms_mgau.c -- Essentially a wrapper that wrap up gauden and
39  * senone. It supports multi-stream.
40  *
41  *
42  * **********************************************
43  * CMU ARPA Speech Project
44  *
45  * Copyright (c) 1997 Carnegie Mellon University.
46  * ALL RIGHTS RESERVED.
47  * **********************************************
48  * HISTORY
49  * $Log$
50  * Revision 1.2 2006/02/22 16:56:01 arthchan2003
51  * Merged from SPHINX3_5_2_RCI_IRII_BRANCH: Added ms_mgau.[ch] into the trunk. It is a wrapper of ms_gauden and ms_senone
52  *
53  * Revision 1.1.2.4 2005/09/25 18:55:19 arthchan2003
54  * Added a flag to turn on and off precomputation.
55  *
56  * Revision 1.1.2.3 2005/08/03 18:53:44 dhdfu
57  * Add memory deallocation functions. Also move all the initialization
58  * of ms_mgau_model_t into ms_mgau_init (duh!), which entails removing it
59  * from decode_anytopo and friends.
60  *
61  * Revision 1.1.2.2 2005/08/02 21:05:38 arthchan2003
62  * 1, Added dist and mgau_active as intermediate variable for computation. 2, Added ms_cont_mgau_frame_eval, which is a multi stream version of GMM computation mainly s3.0 family of tools. 3, Fixed dox-doc.
63  *
64  * Revision 1.1.2.1 2005/07/20 19:37:09 arthchan2003
65  * Added a multi-stream cont_mgau (ms_mgau) which is a wrapper of both gauden and senone. Add ms_mgau_init and model_set_mllr. This allow eliminating 600 lines of code in decode_anytopo/align/allphone.
66  *
67  *
68  *
69  */
70 
71 /* Local headers. */
72 #include "ms_mgau.h"
73 
74 static ps_mgaufuncs_t ms_mgau_funcs = {
75  "ms",
76  &ms_cont_mgau_frame_eval, /* frame_eval */
77  &ms_mgau_mllr_transform, /* transform */
78  &ms_mgau_free /* free */
79 };
80 
81 ps_mgau_t *
82 ms_mgau_init(cmd_ln_t *config, logmath_t *lmath, bin_mdef_t *mdef)
83 {
84  /* Codebooks */
85  ms_mgau_model_t *msg;
86  ps_mgau_t *mg;
87  gauden_t *g;
88  senone_t *s;
89 
90  msg = (ms_mgau_model_t *) ckd_calloc(1, sizeof(ms_mgau_model_t));
91  msg->config = config;
92  msg->g = NULL;
93  msg->s = NULL;
94 
95  g = msg->g = gauden_init(cmd_ln_str_r(config, "-mean"),
96  cmd_ln_str_r(config, "-var"),
97  cmd_ln_float32_r(config, "-varfloor"),
98  lmath);
99  s = msg->s = senone_init(msg->g,
100  cmd_ln_str_r(config, "-mixw"),
101  cmd_ln_str_r(config, "-senmgau"),
102  cmd_ln_float32_r(config, "-mixwfloor"),
103  lmath, mdef);
104 
105  s->aw = cmd_ln_int32_r(config, "-aw");
106 
107  /* Verify senone parameters against gauden parameters */
108  if (s->n_feat != g->n_feat)
109  E_FATAL("#Feature mismatch: gauden= %d, senone= %d\n", g->n_feat,
110  s->n_feat);
111  if (s->n_cw != g->n_density)
112  E_FATAL("#Densities mismatch: gauden= %d, senone= %d\n",
113  g->n_density, s->n_cw);
114  if (s->n_gauden > g->n_mgau)
115  E_FATAL("Senones need more codebooks (%d) than present (%d)\n",
116  s->n_gauden, g->n_mgau);
117  if (s->n_gauden < g->n_mgau)
118  E_ERROR("Senones use fewer codebooks (%d) than present (%d)\n",
119  s->n_gauden, g->n_mgau);
120 
121  msg->topn = cmd_ln_int32_r(config, "-topn");
122  E_INFO("The value of topn: %d\n", msg->topn);
123  if (msg->topn == 0 || msg->topn > msg->g->n_density) {
124  E_WARN
125  ("-topn argument (%d) invalid or > #density codewords (%d); set to latter\n",
126  msg->topn, msg->g->n_density);
127  msg->topn = msg->g->n_density;
128  }
129 
130  msg->dist = (gauden_dist_t ***)
131  ckd_calloc_3d(g->n_mgau, g->n_feat, msg->topn,
132  sizeof(gauden_dist_t));
133  msg->mgau_active = ckd_calloc(g->n_mgau, sizeof(int8));
134 
135  mg = (ps_mgau_t *)msg;
136  mg->vt = &ms_mgau_funcs;
137  return mg;
138 }
139 
140 void
141 ms_mgau_free(ps_mgau_t * mg)
142 {
143  ms_mgau_model_t *msg = (ms_mgau_model_t *)mg;
144  if (msg == NULL)
145  return;
146 
147  gauden_free(msg->g);
148  senone_free(msg->s);
149  ckd_free_3d((void *) msg->dist);
150  ckd_free(msg->mgau_active);
151  ckd_free(msg);
152 }
153 
154 int
155 ms_mgau_mllr_transform(ps_mgau_t *s,
156  ps_mllr_t *mllr)
157 {
158  ms_mgau_model_t *msg = (ms_mgau_model_t *)s;
159  return gauden_mllr_transform(msg->g, mllr, msg->config);
160 }
161 
162 int32
163 ms_cont_mgau_frame_eval(ps_mgau_t * mg,
164  int16 *senscr,
165  uint8 *senone_active,
166  int32 n_senone_active,
167  mfcc_t ** feat,
168  int32 frame,
169  int32 compallsen)
170 {
171  ms_mgau_model_t *msg = (ms_mgau_model_t *)mg;
172  int32 gid;
173  int32 topn;
174  int32 best;
175  gauden_t *g;
176  senone_t *sen;
177 
178  topn = ms_mgau_topn(msg);
179  g = ms_mgau_gauden(msg);
180  sen = ms_mgau_senone(msg);
181 
182  if (compallsen) {
183  int32 s;
184 
185  for (gid = 0; gid < g->n_mgau; gid++)
186  gauden_dist(g, gid, topn, feat, msg->dist[gid]);
187 
188  best = (int32) 0x7fffffff;
189  for (s = 0; s < sen->n_sen; s++) {
190  senscr[s] = senone_eval(sen, s, msg->dist[sen->mgau[s]], topn);
191  if (best > senscr[s]) {
192  best = senscr[s];
193  }
194  }
195 
196  /* Normalize senone scores */
197  for (s = 0; s < sen->n_sen; s++) {
198  int32 bs = senscr[s] - best;
199  if (bs > 32767)
200  bs = 32767;
201  if (bs < -32768)
202  bs = -32768;
203  senscr[s] = bs;
204  }
205  }
206  else {
207  int32 i, n;
208  /* Flag all active mixture-gaussian codebooks */
209  for (gid = 0; gid < g->n_mgau; gid++)
210  msg->mgau_active[gid] = 0;
211 
212  n = 0;
213  for (i = 0; i < n_senone_active; i++) {
214  /* senone_active consists of deltas. */
215  int32 s = senone_active[i] + n;
216  msg->mgau_active[sen->mgau[s]] = 1;
217  n = s;
218  }
219 
220  /* Compute topn gaussian density values (for active codebooks) */
221  for (gid = 0; gid < g->n_mgau; gid++) {
222  if (msg->mgau_active[gid])
223  gauden_dist(g, gid, topn, feat, msg->dist[gid]);
224  }
225 
226  best = (int32) 0x7fffffff;
227  n = 0;
228  for (i = 0; i < n_senone_active; i++) {
229  int32 s = senone_active[i] + n;
230  senscr[s] = senone_eval(sen, s, msg->dist[sen->mgau[s]], topn);
231  if (best > senscr[s]) {
232  best = senscr[s];
233  }
234  n = s;
235  }
236 
237  /* Normalize senone scores */
238  n = 0;
239  for (i = 0; i < n_senone_active; i++) {
240  int32 s = senone_active[i] + n;
241  int32 bs = senscr[s] - best;
242  if (bs > 32767)
243  bs = 32767;
244  if (bs < -32768)
245  bs = -32768;
246  senscr[s] = bs;
247  n = s;
248  }
249  }
250 
251  return 0;
252 }