/**
* <JA>
* 発話検証・棄却用の1状態 GMM を読み込んで初期化する.
*
* </JA>
* <EN>
* Read and initialize an 1-state GMM for utterance verification and
* rejection.
*
* </EN>
*
* @param jconf [in] global configuration variables
*
* @return the newly created GMM information structure in HMM format,
* or NULL on failure.
*/
static HTK_HMM_INFO *
initialize_GMM(Jconf *jconf)
{
HTK_HMM_INFO *gmm;
jlog("STAT: reading GMM: %s\n", jconf->reject.gmm_filename);
if (jconf->gmm == NULL) {
/* no acoustic parameter setting was given for GMM using -AM_GMM,
copy the first AM setting */
jlog("STAT: -AM_GMM not used, use parameter of the first AM\n");
jconf->gmm = j_jconf_am_new();
memcpy(jconf->gmm, jconf->am_root, sizeof(JCONF_AM));
jconf->gmm->hmmfilename = NULL;
jconf->gmm->mapfilename = NULL;
jconf->gmm->spmodel_name = NULL;
jconf->gmm->hmm_gs_filename = NULL;
if (jconf->am_root->analysis.cmnload_filename) {
jconf->gmm->analysis.cmnload_filename = strcpy((char *)mymalloc(strlen(jconf->am_root->analysis.cmnload_filename)+ 1), jconf->am_root->analysis.cmnload_filename);
}
if (jconf->am_root->analysis.cmnsave_filename) {
jconf->gmm->analysis.cmnsave_filename = strcpy((char *)mymalloc(strlen(jconf->am_root->analysis.cmnsave_filename)+ 1), jconf->am_root->analysis.cmnsave_filename);
}
if (jconf->am_root->frontend.ssload_filename) {
jconf->gmm->frontend.ssload_filename = strcpy((char *)mymalloc(strlen(jconf->am_root->frontend.ssload_filename)+ 1), jconf->am_root->frontend.ssload_filename);
}
}
gmm = hmminfo_new();
if (init_hmminfo(gmm, jconf->reject.gmm_filename, NULL, &(jconf->gmm->analysis.para_hmm)) == FALSE) {
hmminfo_free(gmm);
return NULL;
}
/* check parameter type of this acoustic HMM */
if (jconf->input.type == INPUT_WAVEFORM) {
/* Decode parameter extraction type according to the training
parameter type in the header of the given acoustic HMM */
switch(gmm->opt.param_type & F_BASEMASK) {
case F_MFCC:
case F_FBANK:
case F_MELSPEC:
break;
default:
jlog("ERROR: m_fusion: for direct speech input, only GMM trained by MFCC ior filterbank is supported\n");
hmminfo_free(gmm);
return NULL;
}
}
/* set acoustic analysis parameters from HMM header */
calc_para_from_header(&(jconf->gmm->analysis.para), gmm->opt.param_type, gmm->opt.vec_size);
if (jconf->gmm->analysis.para_htk.loaded == 1) apply_para(&(jconf->gmm->analysis.para), &(jconf->gmm->analysis.para_htk));
if (jconf->gmm->analysis.para_hmm.loaded == 1) apply_para(&(jconf->gmm->analysis.para), &(jconf->gmm->analysis.para_hmm));
apply_para(&(jconf->gmm->analysis.para), &(jconf->gmm->analysis.para_default));
return(gmm);
}
/**
* <JA>
* Gaussian Mixture Selection のための状態選択用モノフォンHMMを読み込む.
* </JA>
* <EN>
* Initialize context-independent HMM for state selection with Gaussian
* Mixture Selection.
* </EN>
*
* @param amconf [in] AM configuratino variables
*
* @return the newly created HMM information structure, or NULL on failure.
*/
static HTK_HMM_INFO *
initialize_GSHMM(JCONF_AM *amconf)
{
HTK_HMM_INFO *hmm_gs;
Value para_dummy;
jlog("STAT: Reading GS HMMs:\n");
hmm_gs = hmminfo_new();
undef_para(¶_dummy);
if (init_hmminfo(hmm_gs, amconf->hmm_gs_filename, NULL, ¶_dummy) == FALSE) {
hmminfo_free(hmm_gs);
return NULL;
}
return(hmm_gs);
}
int
main(int argc, char *argv[])
{
FILE *fp;
char *hmmdefs_file;
char *hmmlist_file;
char *outfile;
int i;
hmmdefs_file = hmmlist_file = outfile = NULL;
for(i=1;i<argc;i++) {
if (hmmdefs_file == NULL) {
hmmdefs_file = argv[i];
} else if (hmmlist_file == NULL) {
hmmlist_file = argv[i];
} else if (outfile == NULL) {
outfile = argv[i];
} else {
usage(argv[0]);
return -1;
}
}
if (hmmdefs_file == NULL || hmmlist_file == NULL || outfile == NULL) {
usage(argv[0]);
return -1;
}
hmminfo = hmminfo_new();
printf("---- reading hmmdefs ----\n");
printf("filename: %s\n", hmmdefs_file);
/* read hmmdef file */
undef_para(¶);
if (init_hmminfo(hmminfo, hmmdefs_file, hmmlist_file, ¶) == FALSE) {
fprintf(stderr, "--- terminated\n");
return -1;
}
if (hmminfo->is_triphone) {
fprintf(stderr, "making pseudo bi/mono-phone for IW-triphone\n");
if (make_cdset(hmminfo) == FALSE) {
fprintf(stderr, "ERROR: m_fusion: failed to make context-dependent state set\n");
return -1;
}
}
printf("\n------------------------------------------------------------\n");
print_hmmdef_info(stdout, hmminfo);
printf("\n");
printf("------------------------------------------------------------\n");
printf("---- writing logical-to-physical mapping and pseudo phone info ----\n");
printf("filename: %s\n", outfile);
if ((fp = fopen_writefile(outfile)) == NULL) {
fprintf(stderr, "failed to open %s for writing\n", outfile);
return -1;
}
if (save_hmmlist_bin(fp, hmminfo) == FALSE) {
fprintf(stderr, "failed to write to %s\n", outfile);
return -1;
}
if (fclose_writefile(fp) != 0) {
fprintf(stderr, "failed to close %s\n", outfile);
return -1;
}
printf("\n");
printf("binary HMMList and pseudo phone definitions are written to \"%s\"\n", outfile);
return 0;
}
/**
* <JA>
* @brief 音響HMMをファイルから読み込み，認識用にセットアップする.
*
* ファイルからのHMM定義の読み込み，HMMList ファイルの読み込み，
* パラメータ型のチェック，マルチパス扱いの on/off, ポーズモデルの設定など
* が行われ，認識のための準備が行われる.
*
* この音響モデルの入力となる音響パラメータの種類やパラメータもここで
* 最終決定される. 決定には，音響HMMのヘッダ，（バイナリHMMの場合，存
* 在すれば）バイナリHMMに埋め込まれた特徴量情報，jconf の設定（ばらば
* らに，あるいは -htkconf 使用時）などの情報が用いられる.
* </JA>
* <EN>
* @brief Read in an acoustic HMM from file and setup for recognition.
*
* This functions reads HMM definitions from file, reads also a
* HMMList file, makes logical-to-physical model mapping, determine
* required parameter type, determine whether multi-path handling is needed,
* and find pause model in the definitions.
*
* The feature vector extraction parameters are also finally
* determined in this function. Informations used for the
* determination is (1) the header values in hmmdefs, (2) embedded
* parameters in binary HMM if you are reading a binary HMM made with
* recent mkbinhmm, (3) user-specified parameters in jconf
* configurations (either by separatedly specified or by -htkconf
* options).
*
* </EN>
*
* @param amconf [in] AM configuration variables
* @param jconf [i/o] global configuration variables
*
* @return the newly created HMM information structure, or NULL on failure.
*
*/
static HTK_HMM_INFO *
initialize_HMM(JCONF_AM *amconf, Jconf *jconf)
{
HTK_HMM_INFO *hmminfo;
/* at here, global variable "para" holds values specified by user or
by user-specified HTK config file */
if (amconf->analysis.para_hmm.loaded == 1) {
jlog("Warning: you seems to read more than one acoustic model for recognition, but\n");
jlog("Warning: previous one already has header-embedded acoustic parameters\n");
jlog("Warning: if you have different parameters, result may be wrong!\n");
}
/* allocate new hmminfo */
hmminfo = hmminfo_new();
/* load hmmdefs */
if (init_hmminfo(hmminfo, amconf->hmmfilename, amconf->mapfilename, &(amconf->analysis.para_hmm)) == FALSE) {
hmminfo_free(hmminfo);
return NULL;
}
/* set multipath mode flag */
if (amconf->force_multipath) {
jlog("STAT: m_fusion: force multipath HMM handling by user request\n");
hmminfo->multipath = TRUE;
} else {
hmminfo->multipath = hmminfo->need_multipath;
}
/* only MFCC is supported for audio input */
/* MFCC_{0|E}[_D][_A][_Z][_N] is supported */
/* check parameter type of this acoustic HMM */
if (jconf->input.type == INPUT_WAVEFORM) {
/* Decode parameter extraction type according to the training
parameter type in the header of the given acoustic HMM */
if ((hmminfo->opt.param_type & F_BASEMASK) != F_MFCC) {
jlog("ERROR: m_fusion: for direct speech input, only HMM trained by MFCC is supported\n");
hmminfo_free(hmminfo);
return NULL;
}
/* set acoustic analysis parameters from HMM header */
calc_para_from_header(&(amconf->analysis.para), hmminfo->opt.param_type, hmminfo->opt.vec_size);
}
/* check if tied_mixture */
if (hmminfo->is_tied_mixture && hmminfo->codebooknum <= 0) {
jlog("ERROR: m_fusion: this tied-mixture model has no codebook!?\n");
hmminfo_free(hmminfo);
return NULL;
}
#ifdef PASS1_IWCD
/* make state clusters of same context for inter-word triphone approx. */
if (hmminfo->is_triphone) {
if (hmminfo->cdset_root == NULL) {
jlog("STAT: making pseudo bi/mono-phone for IW-triphone\n");
if (make_cdset(hmminfo) == FALSE) {
jlog("ERROR: m_fusion: failed to make context-dependent state set\n");
hmminfo_free(hmminfo);
return NULL;
}
} else {
jlog("STAT: pseudo phones are loaded from binary hmmlist file\n");
}
/* add those `pseudo' biphone and monophone to the logical HMM names */
/* they points not to the defined HMM, but to the CD_Set structure */
hmm_add_pseudo_phones(hmminfo);
}
#endif
/* find short pause model and set to hmminfo->sp */
htk_hmm_set_pause_model(hmminfo, amconf->spmodel_name);
hmminfo->cdset_method = amconf->iwcdmethod;
hmminfo->cdmax_num = amconf->iwcdmaxn;
if (amconf->analysis.para_htk.loaded == 1) apply_para(&(amconf->analysis.para), &(amconf->analysis.para_htk));
if (amconf->analysis.para_hmm.loaded == 1) apply_para(&(amconf->analysis.para), &(amconf->analysis.para_hmm));
apply_para(&(amconf->analysis.para), &(amconf->analysis.para_default));
return(hmminfo);
}
int
main(int argc, char *argv[])
{
FILE *fp;
char *infile;
char *outfile;
char *conffile;
int i;
infile = outfile = conffile = NULL;
for(i=1;i<argc;i++) {
if (strmatch(argv[i], "-C") || strmatch(argv[i], "-htkconf")) {
if (++i >= argc) {
usage(argv[0]);
return -1;
}
conffile = argv[i];
} else {
if (infile == NULL) {
infile = argv[i];
} else if (outfile == NULL) {
outfile = argv[i];
} else {
usage(argv[0]);
return -1;
}
}
}
if (infile == NULL || outfile == NULL) {
usage(argv[0]);
return -1;
}
hmminfo = hmminfo_new();
printf("---- reading hmmdefs ----\n");
printf("filename: %s\n", infile);
/* read hmmdef file */
undef_para(¶);
if (init_hmminfo(hmminfo, infile, NULL, ¶) == FALSE) {
fprintf(stderr, "--- terminated\n");
return -1;
}
if (conffile != NULL) {
/* if input HMMDEFS already has embedded parameter
they will be overridden by the parameters in the config file */
printf("\n---- reading HTK Config ----\n");
if (para.loaded == 1) {
printf("Warning: input hmmdefs has acoustic analysis parameter information\n");
printf("Warning: they are overridden by the HTK Config file...\n");
}
/* load HTK config file */
undef_para(¶);
if (htk_config_file_parse(conffile, ¶) == FALSE) {
fprintf(stderr, "Error: failed to read %s\n", conffile);
return(-1);
}
/* set some parameters from HTK HMM header information */
printf("\nsetting TARGETKIND and NUMCEPS from HMM definition header...");
calc_para_from_header(¶, hmminfo->opt.param_type, hmminfo->opt.vec_size);
printf("done\n");
/* fulfill unspecified values with HTK defaults */
printf("fulfill unspecified values with HTK defaults...");
undef_para(¶_htk);
make_default_para_htk(¶_htk);
apply_para(¶, ¶_htk);
printf("done\n");
}
printf("\n------------------------------------------------------------\n");
print_hmmdef_info(stdout, hmminfo);
printf("\n");
if (para.loaded == 1) {
put_para(stdout, ¶);
}
printf("------------------------------------------------------------\n");
printf("---- writing ----\n");
printf("filename: %s\n", outfile);
if ((fp = fopen_writefile(outfile)) == NULL) {
fprintf(stderr, "failed to open %s for writing\n", outfile);
return -1;
}
if (write_binhmm(fp, hmminfo, (para.loaded == 1) ? ¶ : NULL) == FALSE) {
fprintf(stderr, "failed to write to %s\n", outfile);
return -1;
}
if (fclose_writefile(fp) != 0) {
fprintf(stderr, "failed to close %s\n", outfile);
return -1;
}
printf("\n");
if (para.loaded == 1) {
printf("binary HMM written to \"%s\", with acoustic parameters embedded for Julius.\n", outfile);
} else {
printf("binary HMM written to \"%s\"\n", outfile);
}
return 0;
}