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 #include <string.h>
00039 #include <stdio.h>
00040 #include <assert.h>
00041
00042
00043 #include <bio.h>
00044
00045
00046 #include "ms_senone.h"
00047
00048
00049 #define MIXW_PARAM_VERSION "1.0"
00050 #define SPDEF_PARAM_VERSION "1.2"
00051
00052 #if defined(__STDC_VERSION__) && (__STDC_VERSION__ == 199901L)
00053 #define LOGMATH_INLINE inline
00054 #elif defined(__GNUC__)
00055 #define LOGMATH_INLINE static inline
00056 #elif defined(_MSC_VER)
00057 #define LOGMATH_INLINE __inline
00058 #else
00059 #define LOGMATH_INLINE static
00060 #endif
00061
00062 static int32
00063 senone_mgau_map_read(senone_t * s, char const *file_name)
00064 {
00065 FILE *fp;
00066 int32 byteswap, chksum_present, n_gauden_present;
00067 uint32 chksum;
00068 int32 i;
00069 char eofchk;
00070 char **argname, **argval;
00071 void *ptr;
00072 float32 v;
00073
00074 E_INFO("Reading senone gauden-codebook map file: %s\n", file_name);
00075
00076 if ((fp = fopen(file_name, "rb")) == NULL)
00077 E_FATAL_SYSTEM("fopen(%s,rb) failed\n", file_name);
00078
00079
00080 if (bio_readhdr(fp, &argname, &argval, &byteswap) < 0)
00081 E_FATAL("bio_readhdr(%s) failed\n", file_name);
00082
00083
00084 chksum_present = 0;
00085 n_gauden_present = 0;
00086 for (i = 0; argname[i]; i++) {
00087 if (strcmp(argname[i], "version") == 0) {
00088 if (strcmp(argval[i], SPDEF_PARAM_VERSION) != 0) {
00089 E_WARN("Version mismatch(%s): %s, expecting %s\n",
00090 file_name, argval[i], SPDEF_PARAM_VERSION);
00091 }
00092
00093
00094 if (sscanf(argval[i], "%f", &v) != 1)
00095 E_FATAL("%s: Bad version no. string: %s\n", file_name,
00096 argval[i]);
00097
00098 n_gauden_present = (v > 1.1) ? 1 : 0;
00099 }
00100 else if (strcmp(argname[i], "chksum0") == 0) {
00101 chksum_present = 1;
00102 }
00103 }
00104 bio_hdrarg_free(argname, argval);
00105 argname = argval = NULL;
00106
00107 chksum = 0;
00108
00109
00110 if (n_gauden_present) {
00111 if (bio_fread
00112 (&(s->n_gauden), sizeof(int32), 1, fp, byteswap, &chksum) != 1)
00113 E_FATAL("fread(%s) (#gauden) failed\n", file_name);
00114 }
00115
00116
00117 if (bio_fread_1d(&ptr, sizeof(int16), &(s->n_sen), fp,
00118 byteswap, &chksum) < 0) {
00119 E_FATAL("bio_fread_1d(%s) failed\n", file_name);
00120 }
00121 s->mgau = ptr;
00122
00123
00124 if (!n_gauden_present) {
00125 s->n_gauden = 1;
00126 for (i = 0; i < s->n_sen; i++)
00127 if (s->mgau[i] >= s->n_gauden)
00128 s->n_gauden = s->mgau[i] + 1;
00129 }
00130
00131 if (chksum_present)
00132 bio_verify_chksum(fp, byteswap, chksum);
00133
00134 if (fread(&eofchk, 1, 1, fp) == 1)
00135 E_FATAL("More data than expected in %s\n", file_name);
00136
00137 fclose(fp);
00138
00139 E_INFO("Read %d->%d senone-codebook mappings\n", s->n_sen,
00140 s->n_gauden);
00141
00142 return 1;
00143 }
00144
00145
00146 static int32
00147 senone_mixw_read(senone_t * s, char const *file_name, logmath_t *lmath)
00148 {
00149 char eofchk;
00150 FILE *fp;
00151 int32 byteswap, chksum_present;
00152 uint32 chksum;
00153 float32 *pdf;
00154 int32 i, f, c, p, n_err;
00155 char **argname, **argval;
00156
00157 E_INFO("Reading senone mixture weights: %s\n", file_name);
00158
00159 if ((fp = fopen(file_name, "rb")) == NULL)
00160 E_FATAL_SYSTEM("fopen(%s,rb) failed\n", file_name);
00161
00162
00163 if (bio_readhdr(fp, &argname, &argval, &byteswap) < 0)
00164 E_FATAL("bio_readhdr(%s) failed\n", file_name);
00165
00166
00167 chksum_present = 0;
00168 for (i = 0; argname[i]; i++) {
00169 if (strcmp(argname[i], "version") == 0) {
00170 if (strcmp(argval[i], MIXW_PARAM_VERSION) != 0)
00171 E_WARN("Version mismatch(%s): %s, expecting %s\n",
00172 file_name, argval[i], MIXW_PARAM_VERSION);
00173 }
00174 else if (strcmp(argname[i], "chksum0") == 0) {
00175 chksum_present = 1;
00176 }
00177 }
00178 bio_hdrarg_free(argname, argval);
00179 argname = argval = NULL;
00180
00181 chksum = 0;
00182
00183
00184 if ((bio_fread(&(s->n_sen), sizeof(int32), 1, fp, byteswap, &chksum) !=
00185 1)
00186 ||
00187 (bio_fread(&(s->n_feat), sizeof(int32), 1, fp, byteswap, &chksum)
00188 != 1)
00189 || (bio_fread(&(s->n_cw), sizeof(int32), 1, fp, byteswap, &chksum)
00190 != 1)
00191 || (bio_fread(&i, sizeof(int32), 1, fp, byteswap, &chksum) != 1)) {
00192 E_FATAL("bio_fread(%s) (arraysize) failed\n", file_name);
00193 }
00194 if (i != s->n_sen * s->n_feat * s->n_cw) {
00195 E_FATAL
00196 ("%s: #float32s(%d) doesn't match dimensions: %d x %d x %d\n",
00197 file_name, i, s->n_sen, s->n_feat, s->n_cw);
00198 }
00199
00200
00201
00202
00203
00204 if ((s->mixwfloor <= 0.0) || (s->mixwfloor >= 1.0))
00205 E_FATAL("mixwfloor (%e) not in range (0, 1)\n", s->mixwfloor);
00206
00207
00208 E_INFO("Truncating senone logs3(pdf) values by %d bits\n", SENSCR_SHIFT);
00209
00210
00211
00212
00213
00214 if (s->n_gauden > 1) {
00215 s->pdf =
00216 (senprob_t ***) ckd_calloc_3d(s->n_sen, s->n_feat, s->n_cw,
00217 sizeof(senprob_t));
00218 }
00219 else {
00220 s->pdf =
00221 (senprob_t ***) ckd_calloc_3d(s->n_feat, s->n_cw, s->n_sen,
00222 sizeof(senprob_t));
00223 }
00224
00225
00226 pdf = (float32 *) ckd_calloc(s->n_cw, sizeof(float32));
00227
00228
00229 n_err = 0;
00230 for (i = 0; i < s->n_sen; i++) {
00231 for (f = 0; f < s->n_feat; f++) {
00232 if (bio_fread
00233 ((void *) pdf, sizeof(float32), s->n_cw, fp, byteswap,
00234 &chksum)
00235 != s->n_cw) {
00236 E_FATAL("bio_fread(%s) (arraydata) failed\n", file_name);
00237 }
00238
00239
00240 if (vector_sum_norm(pdf, s->n_cw) <= 0.0)
00241 n_err++;
00242 vector_floor(pdf, s->n_cw, s->mixwfloor);
00243 vector_sum_norm(pdf, s->n_cw);
00244
00245
00246 for (c = 0; c < s->n_cw; c++) {
00247 p = -(logmath_log(lmath, pdf[c]));
00248 p += (1 << (SENSCR_SHIFT - 1)) - 1;
00249
00250 if (s->n_gauden > 1)
00251 s->pdf[i][f][c] =
00252 (p < (255 << SENSCR_SHIFT)) ? (p >> SENSCR_SHIFT) : 255;
00253 else
00254 s->pdf[f][c][i] =
00255 (p < (255 << SENSCR_SHIFT)) ? (p >> SENSCR_SHIFT) : 255;
00256 }
00257 }
00258 }
00259 if (n_err > 0)
00260 E_ERROR("Weight normalization failed for %d senones\n", n_err);
00261
00262 ckd_free(pdf);
00263
00264 if (chksum_present)
00265 bio_verify_chksum(fp, byteswap, chksum);
00266
00267 if (fread(&eofchk, 1, 1, fp) == 1)
00268 E_FATAL("More data than expected in %s\n", file_name);
00269
00270 fclose(fp);
00271
00272 E_INFO
00273 ("Read mixture weights for %d senones: %d features x %d codewords\n",
00274 s->n_sen, s->n_feat, s->n_cw);
00275
00276 return 1;
00277 }
00278
00279
00280 senone_t *
00281 senone_init(gauden_t *g, char const *mixwfile, char const *sen2mgau_map_file,
00282 float32 mixwfloor, logmath_t *lmath)
00283 {
00284 senone_t *s;
00285 int32 n = 0, i;
00286
00287 s = (senone_t *) ckd_calloc(1, sizeof(senone_t));
00288 s->lmath = logmath_init(logmath_get_base(lmath), SENSCR_SHIFT, TRUE);
00289 s->mixwfloor = mixwfloor;
00290
00291 s->n_gauden = g->n_mgau;
00292 if (sen2mgau_map_file) {
00293 if (!(strcmp(sen2mgau_map_file, ".semi.") == 0
00294 || strcmp(sen2mgau_map_file, ".cont.") == 0)) {
00295 senone_mgau_map_read(s, sen2mgau_map_file);
00296 n = s->n_sen;
00297 }
00298 }
00299 else {
00300 if (s->n_gauden == 1)
00301 sen2mgau_map_file = ".semi.";
00302 else
00303 sen2mgau_map_file = ".cont.";
00304 }
00305
00306 senone_mixw_read(s, mixwfile, lmath);
00307
00308 if (strcmp(sen2mgau_map_file, ".semi.") == 0) {
00309
00310 s->mgau = (int16 *) ckd_calloc(s->n_sen, sizeof(*s->mgau));
00311 }
00312 else if (strcmp(sen2mgau_map_file, ".cont.") == 0
00313 || strcmp(sen2mgau_map_file, ".s3cont.") == 0) {
00314
00315 if (s->n_sen <= 1)
00316 E_FATAL("#senone=%d; must be >1\n", s->n_sen);
00317
00318 s->mgau = (int16 *) ckd_calloc(s->n_sen, sizeof(*s->mgau));
00319 for (i = 0; i < s->n_sen; i++)
00320 s->mgau[i] = i;
00321
00322 s->n_gauden = s->n_sen;
00323 }
00324 else {
00325 if (s->n_sen != n)
00326 E_FATAL("#senones inconsistent: %d in %s; %d in %s\n",
00327 n, sen2mgau_map_file, s->n_sen, mixwfile);
00328 }
00329
00330 s->featscr = NULL;
00331 return s;
00332 }
00333
00334 void
00335 senone_free(senone_t * s)
00336 {
00337 if (s == NULL)
00338 return;
00339 if (s->pdf)
00340 ckd_free_3d((void *) s->pdf);
00341 if (s->mgau)
00342 ckd_free(s->mgau);
00343 if (s->featscr)
00344 ckd_free(s->featscr);
00345 logmath_free(s->lmath);
00346 ckd_free(s);
00347 }
00348
00349
00350
00351
00352
00353
00354
00355 int32
00356 senone_eval(senone_t * s, int id, gauden_dist_t ** dist, int32 n_top)
00357 {
00358 int32 scr;
00359 int32 fden;
00360 int32 fscr;
00361 int32 fwscr;
00362 int32 f, t;
00363 gauden_dist_t *fdist;
00364
00365 assert((id >= 0) && (id < s->n_sen));
00366 assert((n_top > 0) && (n_top <= s->n_cw));
00367
00368 scr = 0;
00369
00370 for (f = 0; f < s->n_feat; f++) {
00371 fdist = dist[f];
00372
00373
00374 fden = ((int32)fdist[0].dist) >> SENSCR_SHIFT;
00375 fscr = (s->n_gauden > 1)
00376 ? (fden - s->pdf[id][f][fdist[0].id])
00377 : (fden - s->pdf[f][fdist[0].id][id]);
00378
00379
00380 for (t = 1; t < n_top; t++) {
00381 fden = ((int32)fdist[t].dist) >> SENSCR_SHIFT;
00382 fwscr = (s->n_gauden > 1) ?
00383 (fden - s->pdf[id][f][fdist[t].id]) :
00384 (fden - s->pdf[f][fdist[t].id][id]);
00385 fscr = logmath_add(s->lmath, fscr, fwscr);
00386 }
00387
00388
00389 scr -= fscr;
00390 }
00391
00392
00393 if (scr > 32767)
00394 scr = 32767;
00395 if (scr < -32768)
00396 scr = -32768;
00397 return scr;
00398 }
00399
00400
00401
00402
00403
00404
00405 void
00406 senone_eval_all(senone_t * s, gauden_dist_t ** dist, int32 n_top,
00407 int16 * senscr)
00408 {
00409 int32 i, f, k, cwdist, scr;
00410
00411 senprob_t *pdf;
00412 int32 *featscr = NULL;
00413 featscr = s->featscr;
00414
00415 assert(s->n_gauden == 1);
00416 assert((n_top > 0) && (n_top <= s->n_cw));
00417
00418 if ((s->n_feat > 1) && (!featscr))
00419 featscr = (int32 *) ckd_calloc(s->n_sen, sizeof(int32));
00420
00421
00422
00423 cwdist = ((int32)dist[0][0].dist) >> SENSCR_SHIFT;
00424 pdf = s->pdf[0][dist[0][0].id];
00425
00426 for (i = 0; i < s->n_sen; i++)
00427 senscr[i] = cwdist - pdf[i];
00428
00429
00430 for (k = 1; k < n_top; k++) {
00431 cwdist = ((int32)dist[0][k].dist) >> SENSCR_SHIFT;
00432 pdf = s->pdf[0][dist[0][k].id];
00433
00434 for (i = 0; i < s->n_sen; i++) {
00435 scr = cwdist - pdf[i];
00436 senscr[i] = logmath_add(s->lmath, senscr[i], scr);
00437 if (k == n_top - 1)
00438 senscr[i] = -senscr[i];
00439 }
00440 }
00441
00442
00443 for (f = 1; f < s->n_feat; f++) {
00444
00445 cwdist = ((int32)dist[f][0].dist) >> SENSCR_SHIFT;
00446 pdf = s->pdf[f][dist[f][0].id];
00447
00448 for (i = 0; i < s->n_sen; i++)
00449 featscr[i] = cwdist - pdf[i];
00450
00451
00452 for (k = 1; k < n_top; k++) {
00453 cwdist = ((int32)dist[f][k].dist) >> SENSCR_SHIFT;
00454 pdf = s->pdf[f][dist[f][k].id];
00455
00456 for (i = 0; i < s->n_sen; i++) {
00457 scr = cwdist - pdf[i];
00458 featscr[i] = logmath_add(s->lmath, featscr[i], scr);
00459 }
00460 }
00461
00462 for (i = 0; i < s->n_sen; i++)
00463 senscr[i] -= featscr[i];
00464 }
00465 }