00001
00002
00003
00004
00005
00006
00007
00008
00009
00010
00011
00012
00013
00014
00015
00016
00017
00018
00019
00020
00021
00022
00023
00024
00025
00026
00027
00028
00029
00030
00031
00032
00033
00034
00035
00036
00037
00038
00039
00040
00041
00042
00043
00044
00045
00046
00047
00048
00049
00050
00051
00052
00053
00054
00055
00056
00057
00058
00059
00060
00061
00062
00063
00064
00065
00066
00067
00068
00069
00070
00071 #include "ms_mgau.h"
00072
00073 ms_mgau_model_t *
00074 ms_mgau_init(cmd_ln_t *config, logmath_t *lmath)
00075 {
00076
00077 int32 i;
00078 ms_mgau_model_t *msg;
00079 gauden_t *g;
00080 senone_t *s;
00081 mgau2sen_t *m2s;
00082
00083 msg = (ms_mgau_model_t *) ckd_calloc(1, sizeof(ms_mgau_model_t));
00084
00085
00086 msg->g = NULL;
00087 msg->s = NULL;
00088
00089 msg->g = gauden_init(cmd_ln_str_r(config, "-mean"),
00090 cmd_ln_str_r(config, "-var"),
00091 cmd_ln_float32_r(config, "-varfloor"),
00092 lmath);
00093 msg->s = senone_init(msg->g,
00094 cmd_ln_str_r(config, "-mixw"), NULL,
00095 cmd_ln_float32_r(config, "-mixwfloor"), lmath);
00096
00097 g = ms_mgau_gauden(msg);
00098 s = ms_mgau_senone(msg);
00099
00100
00101 if (s->n_feat != g->n_feat)
00102 E_FATAL("#Feature mismatch: gauden= %d, senone= %d\n", g->n_feat,
00103 s->n_feat);
00104 if (s->n_cw != g->n_density)
00105 E_FATAL("#Densities mismatch: gauden= %d, senone= %d\n",
00106 g->n_density, s->n_cw);
00107 if (s->n_gauden > g->n_mgau)
00108 E_FATAL("Senones need more codebooks (%d) than present (%d)\n",
00109 s->n_gauden, g->n_mgau);
00110 if (s->n_gauden < g->n_mgau)
00111 E_ERROR("Senones use fewer codebooks (%d) than present (%d)\n",
00112 s->n_gauden, g->n_mgau);
00113
00114 msg->mgau2sen =
00115 (mgau2sen_t **) ckd_calloc(g->n_mgau, sizeof(mgau2sen_t *));
00116 for (i = 0; i < s->n_sen; i++) {
00117 m2s = (mgau2sen_t *) ckd_calloc(1, sizeof(mgau2sen_t));
00118 m2s->sen = i;
00119 m2s->next = msg->mgau2sen[s->mgau[i]];
00120 msg->mgau2sen[s->mgau[i]] = m2s;
00121 }
00122
00123 msg->topn = cmd_ln_int32_r(config, "-topn");
00124 E_INFO("The value of topn: %d\n", msg->topn);
00125 if (msg->topn == 0 || msg->topn > msg->g->n_density) {
00126 E_WARN
00127 ("-topn argument (%d) invalid or > #density codewords (%d); set to latter\n",
00128 msg->topn, msg->g->n_density);
00129 msg->topn = msg->g->n_density;
00130 }
00131
00132 msg->dist = (gauden_dist_t ***)
00133 ckd_calloc_3d(g->n_mgau, g->n_feat, msg->topn,
00134 sizeof(gauden_dist_t));
00135 msg->mgau_active = ckd_calloc(g->n_mgau, sizeof(int8));
00136
00137 return msg;
00138 }
00139
00140 void
00141 ms_mgau_free(ms_mgau_model_t * msg)
00142 {
00143 if (msg == NULL)
00144 return;
00145
00146 gauden_free(msg->g);
00147 senone_free(msg->s);
00148 ckd_free_3d((void *) msg->dist);
00149 ckd_free(msg->mgau_active);
00150 ckd_free(msg);
00151 }
00152
00153 int32
00154 ms_cont_mgau_frame_eval(ms_mgau_model_t * msg,
00155 int16 *senscr,
00156 int32 *senone_active,
00157 int32 n_senone_active,
00158 mfcc_t ** feat,
00159 int32 frame,
00160 int32 compallsen,
00161 int32 *bestidx)
00162 {
00163 int32 gid;
00164 int32 i;
00165 int32 topn;
00166 int32 best;
00167 gauden_t *g;
00168 senone_t *sen;
00169
00170 topn = ms_mgau_topn(msg);
00171 g = ms_mgau_gauden(msg);
00172 sen = ms_mgau_senone(msg);
00173
00174
00175
00176
00177
00178
00179
00180 for (gid = 0; gid < g->n_mgau; gid++)
00181 msg->mgau_active[gid] = 0;
00182
00183 for (i = 0; i < n_senone_active; i++) {
00184 int32 s = senone_active[i];
00185 msg->mgau_active[sen->mgau[s]] = 1;
00186 }
00187
00188
00189 for (gid = 0; gid < g->n_mgau; gid++) {
00190 if (msg->mgau_active[gid])
00191 gauden_dist(g, gid, topn, feat, msg->dist[gid]);
00192 }
00193
00194 best = (int32) 0x7fffffff;
00195 *bestidx = -1;
00196 for (i = 0; i < n_senone_active; i++) {
00197 int32 s = senone_active[i];
00198 senscr[s] = senone_eval(sen, s, msg->dist[sen->mgau[s]], topn);
00199 if (best > senscr[s]) {
00200 best = senscr[s];
00201 *bestidx = s;
00202 }
00203 }
00204
00205
00206 for (i = 0; i < n_senone_active; i++) {
00207 int32 s = senone_active[i];
00208 int32 bs = senscr[s] - best;
00209 if (bs > 32767)
00210 bs = 32767;
00211 if (bs < -32768)
00212 bs = -32768;
00213 senscr[s] = bs;
00214 }
00215
00216 return best;
00217 }