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
00072 #include "ms_mgau.h"
00073
00074 static ps_mgaufuncs_t ms_mgau_funcs = {
00075 "ms",
00076 &ms_cont_mgau_frame_eval,
00077 &ms_mgau_mllr_transform,
00078 &ms_mgau_free
00079 };
00080
00081 ps_mgau_t *
00082 ms_mgau_init(cmd_ln_t *config, logmath_t *lmath, bin_mdef_t *mdef)
00083 {
00084
00085 ms_mgau_model_t *msg;
00086 ps_mgau_t *mg;
00087 gauden_t *g;
00088 senone_t *s;
00089
00090 msg = (ms_mgau_model_t *) ckd_calloc(1, sizeof(ms_mgau_model_t));
00091 msg->config = config;
00092 msg->g = NULL;
00093 msg->s = NULL;
00094
00095 msg->g = gauden_init(cmd_ln_str_r(config, "-mean"),
00096 cmd_ln_str_r(config, "-var"),
00097 cmd_ln_float32_r(config, "-varfloor"),
00098 lmath);
00099 msg->s = senone_init(msg->g,
00100 cmd_ln_str_r(config, "-mixw"),
00101 cmd_ln_str_r(config, "-senmgau"),
00102 cmd_ln_float32_r(config, "-mixwfloor"),
00103 lmath, mdef);
00104
00105 g = ms_mgau_gauden(msg);
00106 s = ms_mgau_senone(msg);
00107
00108
00109 if (s->n_feat != g->n_feat)
00110 E_FATAL("#Feature mismatch: gauden= %d, senone= %d\n", g->n_feat,
00111 s->n_feat);
00112 if (s->n_cw != g->n_density)
00113 E_FATAL("#Densities mismatch: gauden= %d, senone= %d\n",
00114 g->n_density, s->n_cw);
00115 if (s->n_gauden > g->n_mgau)
00116 E_FATAL("Senones need more codebooks (%d) than present (%d)\n",
00117 s->n_gauden, g->n_mgau);
00118 if (s->n_gauden < g->n_mgau)
00119 E_ERROR("Senones use fewer codebooks (%d) than present (%d)\n",
00120 s->n_gauden, g->n_mgau);
00121
00122 msg->topn = cmd_ln_int32_r(config, "-topn");
00123 E_INFO("The value of topn: %d\n", msg->topn);
00124 if (msg->topn == 0 || msg->topn > msg->g->n_density) {
00125 E_WARN
00126 ("-topn argument (%d) invalid or > #density codewords (%d); set to latter\n",
00127 msg->topn, msg->g->n_density);
00128 msg->topn = msg->g->n_density;
00129 }
00130
00131 msg->dist = (gauden_dist_t ***)
00132 ckd_calloc_3d(g->n_mgau, g->n_feat, msg->topn,
00133 sizeof(gauden_dist_t));
00134 msg->mgau_active = ckd_calloc(g->n_mgau, sizeof(int8));
00135
00136 mg = (ps_mgau_t *)msg;
00137 mg->vt = &ms_mgau_funcs;
00138 return mg;
00139 }
00140
00141 void
00142 ms_mgau_free(ps_mgau_t * mg)
00143 {
00144 ms_mgau_model_t *msg = (ms_mgau_model_t *)mg;
00145 if (msg == NULL)
00146 return;
00147
00148 gauden_free(msg->g);
00149 senone_free(msg->s);
00150 ckd_free_3d((void *) msg->dist);
00151 ckd_free(msg->mgau_active);
00152 ckd_free(msg);
00153 }
00154
00155 int
00156 ms_mgau_mllr_transform(ps_mgau_t *s,
00157 ps_mllr_t *mllr)
00158 {
00159 ms_mgau_model_t *msg = (ms_mgau_model_t *)s;
00160 return gauden_mllr_transform(msg->g, mllr, msg->config);
00161 }
00162
00163 int32
00164 ms_cont_mgau_frame_eval(ps_mgau_t * mg,
00165 int16 *senscr,
00166 uint8 *senone_active,
00167 int32 n_senone_active,
00168 mfcc_t ** feat,
00169 int32 frame,
00170 int32 compallsen)
00171 {
00172 ms_mgau_model_t *msg = (ms_mgau_model_t *)mg;
00173 int32 gid;
00174 int32 topn;
00175 int32 best;
00176 gauden_t *g;
00177 senone_t *sen;
00178
00179 topn = ms_mgau_topn(msg);
00180 g = ms_mgau_gauden(msg);
00181 sen = ms_mgau_senone(msg);
00182
00183 if (compallsen) {
00184 int32 s;
00185
00186 for (gid = 0; gid < g->n_mgau; gid++)
00187 gauden_dist(g, gid, topn, feat, msg->dist[gid]);
00188
00189 best = (int32) 0x7fffffff;
00190 for (s = 0; s < sen->n_sen; s++) {
00191 senscr[s] = senone_eval(sen, s, msg->dist[sen->mgau[s]], topn);
00192 if (best > senscr[s]) {
00193 best = senscr[s];
00194 }
00195 }
00196
00197
00198 for (s = 0; s < sen->n_sen; s++) {
00199 int32 bs = senscr[s] - best;
00200 if (bs > 32767)
00201 bs = 32767;
00202 if (bs < -32768)
00203 bs = -32768;
00204 senscr[s] = bs;
00205 }
00206 }
00207 else {
00208 int32 i, n;
00209
00210 for (gid = 0; gid < g->n_mgau; gid++)
00211 msg->mgau_active[gid] = 0;
00212
00213 n = 0;
00214 for (i = 0; i < n_senone_active; i++) {
00215
00216 int32 s = senone_active[i] + n;
00217 msg->mgau_active[sen->mgau[s]] = 1;
00218 n = s;
00219 }
00220
00221
00222 for (gid = 0; gid < g->n_mgau; gid++) {
00223 if (msg->mgau_active[gid])
00224 gauden_dist(g, gid, topn, feat, msg->dist[gid]);
00225 }
00226
00227 best = (int32) 0x7fffffff;
00228 n = 0;
00229 for (i = 0; i < n_senone_active; i++) {
00230 int32 s = senone_active[i] + n;
00231 senscr[s] = senone_eval(sen, s, msg->dist[sen->mgau[s]], topn);
00232 if (best > senscr[s]) {
00233 best = senscr[s];
00234 }
00235 n = s;
00236 }
00237
00238
00239 n = 0;
00240 for (i = 0; i < n_senone_active; i++) {
00241 int32 s = senone_active[i] + n;
00242 int32 bs = senscr[s] - best;
00243 if (bs > 32767)
00244 bs = 32767;
00245 if (bs < -32768)
00246 bs = -32768;
00247 senscr[s] = bs;
00248 n = s;
00249 }
00250 }
00251
00252 return 0;
00253 }