/**
* <EN>
* Allocate a new MFCC calculation instance
* </EN>
* <JA>
* MFCC計算インスタンスを新たに割り付ける.
* </JA>
*
* @param amconf [in] acoustic model configuration parameters
*
* @return the newly allocated MFCC calculation instance.
*
* @callgraph
* @callergraph
*
*/
MFCCCalc *
j_mfcccalc_new(JCONF_AM *amconf)
{
MFCCCalc *mfcc;
mfcc = (MFCCCalc *)mymalloc(sizeof(MFCCCalc));
memset(mfcc, 0, sizeof(MFCCCalc));
mfcc->param = NULL;
mfcc->rest_param = NULL;
mfcc->frontend.ssbuf = NULL;
mfcc->cmn.loaded = FALSE;
mfcc->plugin_source = -1;
if (amconf) {
mfcc->para = &(amconf->analysis.para);
mfcc->hmm_loaded = (amconf->analysis.para_hmm.loaded == 1) ? TRUE : FALSE;
mfcc->htk_loaded = (amconf->analysis.para_htk.loaded == 1) ? TRUE : FALSE;
mfcc->wrk = WMP_work_new(mfcc->para);
if (mfcc->wrk == NULL) {
jlog("ERROR: j_mfcccalc_new: failed to initialize MFCC computation\n");
return NULL;
}
mfcc->cmn.load_filename = amconf->analysis.cmnload_filename;
mfcc->cmn.update = amconf->analysis.cmn_update;
mfcc->cmn.save_filename = amconf->analysis.cmnsave_filename;
mfcc->cmn.map_weight = amconf->analysis.cmn_map_weight;
mfcc->frontend.ss_alpha = amconf->frontend.ss_alpha;
mfcc->frontend.ss_floor = amconf->frontend.ss_floor;
mfcc->frontend.sscalc = amconf->frontend.sscalc;
mfcc->frontend.sscalc_len = amconf->frontend.sscalc_len;
mfcc->frontend.ssload_filename = amconf->frontend.ssload_filename;
}
mfcc->next = NULL;
return mfcc;
}
int
main(int argc, char *argv[])
{
Recog *recog;
Jconf *jconf;
float *ss;
MFCCWork *wrk;
/* create instance */
recog = j_recog_new();
jconf = j_jconf_new();
recog->jconf = jconf;
/* set application-specific additional options */
j_add_option("-freq", 1, 1, "sampling freq in Hz", opt_freq);
j_add_option("-len", 1, 1, "record length in msec", opt_len);
j_add_option("-h", 0, 0, "display this help", opt_help);
j_add_option("-help", 0, 0, "display this help", opt_help);
j_add_option("--help", 0, 0, "display this help", opt_help);
/* when no argument, output help and exit */
if (argc <= 1) {
opt_help(jconf, NULL, 0);
return 0;
}
/* regard last arg as filename */
if (strmatch(argv[argc-1], "-")) {
stout = TRUE;
} else {
filename = argv[argc-1];
}
/* set default as same as "-input mic" */
jconf->input.type = INPUT_WAVEFORM;
jconf->input.speech_input = SP_MIC;
jconf->input.device = SP_INPUT_DEFAULT;
/* process config and load them */
if (j_config_load_args(jconf, argc-1, argv) == -1) {
fprintf(stderr, "Error reading arguments\n");
return -1;
}
/* force some default values */
jconf->detect.silence_cut = 0; /* disable silence cut */
jconf->preprocess.strip_zero_sample = TRUE; /* strip zero samples */
jconf->detect.level_thres = 0; /* no VAD, record all */
/* set Julius default parameters for unspecified acoustic parameters */
apply_para(&(jconf->am_root->analysis.para), &(jconf->am_root->analysis.para_default));
/* set some values */
jconf->input.sfreq = jconf->am_root->analysis.para.smp_freq;
jconf->input.period = jconf->am_root->analysis.para.smp_period;
jconf->input.frameshift = jconf->am_root->analysis.para.frameshift;
jconf->input.framesize = jconf->am_root->analysis.para.framesize;
sfreq = jconf->am_root->analysis.para.smp_freq;
/* output file check */
if (!stout) {
if (access(filename, F_OK) == 0) {
if (access(filename, W_OK) == 0) {
fprintf(stderr, "Warning: overwriting file \"%s\"\n", filename);
} else {
perror("mkss");
return(1);
}
}
}
/* allocate speech store buffer */
samples = sfreq * slen / 1000;
speech = (SP16 *)mymalloc(sizeof(SP16) * samples);
/* allocate work area to compute spectrum */
wrk = WMP_work_new(&(jconf->am_root->analysis.para));
if (wrk == NULL) {
jlog("ERROR: m_fusion: failed to initialize MFCC computation for SS\n");
return -1;
}
/* initialize input device */
if (j_adin_init(recog) == FALSE) {
fprintf(stderr, "Error in initializing adin device\n");
return -1;
}
/* open device */
if (j_open_stream(recog, NULL) < 0) {
fprintf(stderr, "Error in opening adin device\n");
}
/* record mic input */
fprintf(stderr, "%dHz recording for %.2f seconds of noise\n", sfreq, (float)slen /(float)1000);
speechnum = 0;
adin_go(adin_callback, NULL, recog);
/* close device */
adin_end(recog->adin);
fprintf(stderr, "\n%d samples (%d bytes, %.1f sec) recorded\n", samples, samples * sizeof(SP16), (float)samples / (float)sfreq);
/* compute SS */
fprintf(stderr, "compute SS:\n");
fprintf(stderr, " fsize : %4d samples (%.1f msec)\n", jconf->input.framesize, (float)jconf->input.framesize * 1000.0/ (float)sfreq);
fprintf(stderr, " fshift: %4d samples (%.1f msec)\n", jconf->input.frameshift, (float)jconf->input.frameshift * 1000.0/ (float)sfreq);
ss = new_SS_calculate(speech, samples, &sslen, wrk, &(jconf->am_root->analysis.para));
fprintf(stderr, " points: %4d\n", sslen);
fprintf(stderr, "noise spectrum was measured\n");
/* open file for recording */
fprintf(stderr, "writing average noise spectrum to [%s]...", filename);
if (stout) {
fd = 1;
} else {
if ((fd = open(filename, O_CREAT | O_RDWR
#ifdef O_BINARY
| O_BINARY
#endif
, 0644)) == -1) {
perror("mkss");
return(1);
}
}
x = sslen;
#ifndef WORDS_BIGENDIAN
swap_bytes((char *)&x, sizeof(int), 1);
#endif
if (write(fd, &x, sizeof(int)) < sizeof(int)) {
perror("mkss");
return(1);
}
#ifndef WORDS_BIGENDIAN
swap_bytes((char *)ss, sizeof(float), sslen);
#endif
if (write(fd, ss, sslen * sizeof(float)) < sslen * sizeof(float)) {
perror("mkss");
return(1);
}
if (!stout) {
if (close(fd) < 0) {
perror("mkss");
return(1);
}
}
fprintf(stderr, "done\n");
WMP_free(wrk);
return 0;
}
/**
* <EN>
* @brief Combine all loaded models and settings into one engine instance.
*
* This function will finalize preparation of recognition:
*
* - create required MFCC calculation instances,
* - create recognition process instance for specified LM/AM combination,
* - set model-specific recognition parameters,
* - build tree lexicon for each process instance for the 1st pass,
* - prepare work area and cache area for recognition,
* - initialize some values / work area for frontend processing.
*
* After this function, all recognition setup was done and we are ready for
* start recognition.
*
* This should be called after j_jconf_finalize() and j_load_all() has been
* completed. You should put the jconf at recog->jconf before calling this
* function.
* </EN>
* <JA>
* @brief 全てのロードされたモデルと設定からエンジンインスタンスを
* 最終構成する.
*
* この関数は，認識準備のための最終処理を行う. 内部では，
*
* - 必要な MFCC 計算インスタンスの生成
* - 指定された LM/AM の組からの認識処理インスタンス生成
* - モデルに依存する認識用パラメータの設定
* - 第1パス用の木構造化辞書を認識処理インスタンスごとに構築
* - 認識処理用ワークエリアとキャッシュエリアを確保
* - フロントエンド処理のためのいくつかの値とワークエリアの確保
*
* を行う. この関数が終了後，エンジンインスタンス内の全てのセットアップ
* は終了し，認識が開始できる状態となる.
*
* この関数は，j_jconf_finalize() と j_load_all() が終わった状態で
* 呼び出す必要がある. 呼出し前には，recog->jconf に (j_load_all でともに
* 使用した) jconf を格納しておくこと.
*
* </JA>
*
* @param recog [in] engine instance
*
* @return TRUE when all initialization successfully done, or FALSE if any
* error has been occured.
*
* @callgraph
* @callergraph
* @ingroup instance
*
*/
boolean
j_final_fusion(Recog *recog)
{
MFCCCalc *mfcc;
JCONF_SEARCH *sconf;
PROCESS_AM *am;
jlog("STAT: ------\n");
jlog("STAT: All models are ready, go for final fusion\n");
jlog("STAT: [1] create MFCC extraction instance(s)\n");
if (recog->jconf->input.type == INPUT_WAVEFORM) {
/***************************************************/
/* create MFCC calculation instance from AM config */
/* according to the fixated parameter information */
/***************************************************/
create_mfcc_calc_instances(recog);
}
/****************************************/
/* create recognition process instances */
/****************************************/
jlog("STAT: [2] create recognition processing instance(s) with AM and LM\n");
for(sconf=recog->jconf->search_root;sconf;sconf=sconf->next) {
if (j_launch_recognition_instance(recog, sconf) == FALSE) return FALSE;
}
/****************************/
/****** initialize GMM ******/
/****************************/
if (recog->gmm != NULL) {
jlog("STAT: [2.5] create GMM instance\n");
if (gmm_init(recog) == FALSE) {
jlog("ERROR: m_fusion: error in initializing GMM\n");
return FALSE;
}
}
/* stage 4: setup output probability function for each AM */
jlog("STAT: [3] initialize for acoustic HMM calculation\n");
for(am=recog->amlist;am;am=am->next) {
#ifdef ENABLE_PLUGIN
/* set plugin function if specified */
if (am->config->gprune_method == GPRUNE_SEL_USER) {
am->hmmwrk.compute_gaussset = (void (*)(HMMWork *, HTK_HMM_Dens **, int, int *, int)) plugin_get_func(am->config->gprune_plugin_source, "calcmix");
if (am->hmmwrk.compute_gaussset == NULL) {
jlog("ERROR: calcmix plugin has no function \"calcmix\"\n");
return FALSE;
}
am->hmmwrk.compute_gaussset_init = (boolean (*)(HMMWork *)) plugin_get_func(am->config->gprune_plugin_source, "calcmix_init");
if (am->hmmwrk.compute_gaussset_init == NULL) {
jlog("ERROR: calcmix plugin has no function \"calcmix_init\"\n");
return FALSE;
}
am->hmmwrk.compute_gaussset_free = (void (*)(HMMWork *)) plugin_get_func(am->config->gprune_plugin_source, "calcmix_free");
if (am->hmmwrk.compute_gaussset_free == NULL) {
jlog("ERROR: calcmix plugin has no function \"calcmix_free\"\n");
return FALSE;
}
}
#endif
if (am->config->hmm_gs_filename != NULL) {/* with GMS */
if (outprob_init(&(am->hmmwrk), am->hmminfo, am->hmm_gs, am->config->gs_statenum, am->config->gprune_method, am->config->mixnum_thres) == FALSE) {
return FALSE;
}
} else {
if (outprob_init(&(am->hmmwrk), am->hmminfo, NULL, 0, am->config->gprune_method, am->config->mixnum_thres) == FALSE) {
return FALSE;
}
}
}
/* stage 5: initialize work area for input and realtime decoding */
jlog("STAT: [4] prepare MFCC storage(s)\n");
if (recog->jconf->input.type == INPUT_VECTOR) {
/* create an MFCC instance for MFCC input */
/* create new mfcc instance */
recog->mfcclist = j_mfcccalc_new(NULL);
recog->mfcclist->id = 1;
/* assign to the am */
for(am=recog->amlist;am;am=am->next) {
am->mfcc = recog->mfcclist;
}
if (recog->gmm) recog->gmmmfcc = recog->mfcclist;
}
/* allocate parameter holders */
for(mfcc=recog->mfcclist;mfcc;mfcc=mfcc->next) {
mfcc->param = new_param();
}
/* initialize SS calculation work area */
if (recog->jconf->input.type == INPUT_WAVEFORM) {
for(mfcc=recog->mfcclist;mfcc;mfcc=mfcc->next) {
if (mfcc->frontend.sscalc) {
mfcc->frontend.mfccwrk_ss = WMP_work_new(mfcc->para);
if (mfcc->frontend.mfccwrk_ss == NULL) {
jlog("ERROR: m_fusion: failed to initialize MFCC computation for SS\n");
return FALSE;
}
if (mfcc->frontend.sscalc_len * recog->jconf->input.sfreq / 1000 < mfcc->para->framesize) {
jlog("ERROR: m_fusion: head sil length for SS (%d msec) is shorter than a frame (%d msec)\n", mfcc->frontend.sscalc_len, mfcc->para->framesize * 1000 / recog->jconf->input.sfreq);
return FALSE;
}
}
}
}
if (recog->jconf->decodeopt.realtime_flag) {
jlog("STAT: [5] prepare for real-time decoding\n");
/* prepare for 1st pass pipeline processing */
if (recog->jconf->input.type == INPUT_WAVEFORM) {
if (RealTimeInit(recog) == FALSE) {
jlog("ERROR: m_fusion: failed to initialize recognition process\n");
return FALSE;
}
}
}
/* finished! */
jlog("STAT: All init successfully done\n\n");
/* set-up callback plugin if any */
#ifdef ENABLE_PLUGIN
if (plugin_exec_engine_startup(recog) == FALSE) {
jlog("ERROR: m_fusion: failed to execute callback setup in plugin\n");
return FALSE;
}
#endif
return TRUE;
}
