boolean
mfc_module_set_header(MFCCCalc *mfcc, Recog *recog)
{
FUNC_INT func;
unsigned int ret;
func = (FUNC_INT) plugin_get_func(mfcc->plugin_source, "fvin_get_configuration");
if (func == NULL) {
jlog("ERROR: feature vector input plugin: fvin_get_configuration() not found\n");
return FALSE;
}
/* vector length in unit */
mfcc->param->veclen = (*func)(0);
mfcc->param->header.sampsize = mfcc->param->veclen * sizeof(VECT);
/* frame shift in msec */
mfcc->param->header.wshift = (*func)(1) * 10000.0;
/* parameter type for checking (return 0xffff to disable the check) */
ret = (*func)(2);
if (ret == 0xffff) {
/* disable type checking */
recog->jconf->input.paramtype_check_flag = FALSE;
} else {
mfcc->param->header.samptype = ret;
}
return TRUE;
}
boolean
mfc_module_set_header(MFCCCalc *mfcc, Recog *recog)
{
FUNC_INT func;
unsigned int ret;
#ifdef ENABLE_PLUGIN
if (mfcc->plugin_source < 0) {
/* no plugin, use the default functions */
func = vecin_get_configuration;
} else {
func = (FUNC_INT) plugin_get_func(mfcc->plugin_source, "fvin_get_configuration");
if (func == NULL) {
jlog("ERROR: feature vector input: fvin_get_configuration() not found\n");
return FALSE;
}
}
#else
func = vecin_get_configuration;
#endif
/* vector length in unit */
mfcc->param->veclen = (*func)(0);
mfcc->param->header.sampsize = mfcc->param->veclen * sizeof(VECT);
/* frame shift in msec */
mfcc->param->header.wshift = (*func)(1) * 10000.0;
/* parameter type for checking (return 0xffff to disable the check) */
ret = (*func)(2);
if (ret == 0xffff) {
/* disable type checking */
recog->jconf->input.paramtype_check_flag = FALSE;
} else {
mfcc->param->header.samptype = ret;
}
/* switch if the input vector is feature vector or outprob vector */
mfcc->param->is_outprob = ( (*func)(3) > 0 ) ? TRUE : FALSE;
return TRUE;
}
boolean
mfc_module_init(MFCCCalc *mfcc, Recog *recog)
{
mfcc->plugin_source = recog->jconf->input.plugin_source;
if (mfcc->plugin_source < 0) {
jlog("ERROR: SP_MDCMODULE selected but plugin is missing?\n");
return FALSE;
}
mfcc->func.fv_standby = (boolean (*)()) plugin_get_func(mfcc->plugin_source, "fvin_standby");
mfcc->func.fv_begin = (boolean (*)()) plugin_get_func(mfcc->plugin_source, "fvin_open");
mfcc->func.fv_read = (int (*)(VECT *, int)) plugin_get_func(mfcc->plugin_source, "fvin_read");
mfcc->func.fv_end = (boolean (*)()) plugin_get_func(mfcc->plugin_source, "fvin_close");
mfcc->func.fv_resume = (boolean (*)()) plugin_get_func(mfcc->plugin_source, "fvin_resume");
mfcc->func.fv_pause = (boolean (*)()) plugin_get_func(mfcc->plugin_source, "fvin_pause");
mfcc->func.fv_terminate= (boolean (*)()) plugin_get_func(mfcc->plugin_source, "fvin_terminate");
mfcc->func.fv_input_name= (char * (*)()) plugin_get_func(mfcc->plugin_source, "fvin_input_name");
if (mfcc->func.fv_read == NULL) {
jlog("ERROR: FEATURE_INPUT plugin: fvin_read() not found!\n");
return FALSE;
}
return TRUE;
}
boolean
mfc_module_init(MFCCCalc *mfcc, Recog *recog)
{
/* assign default functions */
mfcc->func.fv_standby = (boolean (*)()) vecin_standby;
mfcc->func.fv_begin = (boolean (*)()) vecin_open;
mfcc->func.fv_read = (int (*)(VECT *, int)) vecin_read;
mfcc->func.fv_end = (boolean (*)()) vecin_close;
mfcc->func.fv_resume = (boolean (*)()) vecin_resume;
mfcc->func.fv_pause = (boolean (*)()) vecin_pause;
mfcc->func.fv_terminate = (boolean (*)()) vecin_terminate;
mfcc->func.fv_input_name = (char * (*)()) vecin_input_name;
#ifdef ENABLE_PLUGIN
mfcc->plugin_source = recog->jconf->input.plugin_source;
if (mfcc->plugin_source < 0) {
/* no plugin, use the default functions */
return TRUE;
}
mfcc->func.fv_standby = (boolean (*)()) plugin_get_func(mfcc->plugin_source, "fvin_standby");
mfcc->func.fv_begin = (boolean (*)()) plugin_get_func(mfcc->plugin_source, "fvin_open");
mfcc->func.fv_read = (int (*)(VECT *, int)) plugin_get_func(mfcc->plugin_source, "fvin_read");
mfcc->func.fv_end = (boolean (*)()) plugin_get_func(mfcc->plugin_source, "fvin_close");
mfcc->func.fv_resume = (boolean (*)()) plugin_get_func(mfcc->plugin_source, "fvin_resume");
mfcc->func.fv_pause = (boolean (*)()) plugin_get_func(mfcc->plugin_source, "fvin_pause");
mfcc->func.fv_terminate= (boolean (*)()) plugin_get_func(mfcc->plugin_source, "fvin_terminate");
mfcc->func.fv_input_name= (char * (*)()) plugin_get_func(mfcc->plugin_source, "fvin_input_name");
if (mfcc->func.fv_read == NULL) {
jlog("ERROR: FEATURE_INPUT: fvin_read() not found!\n");
return FALSE;
}
#endif
return TRUE;
}
/**
* <JA>
* 設定パラメータに従い音声入力デバイスをセットアップする.
*
* @param recog [i/o] エンジンインスタンス
*
* </JA>
* <EN>
* Set up audio input device according to the jconf configurations.
*
* @param recog [i/o] engine instance
* </EN>
*
* @callgraph
* @callergraph
*/
boolean
adin_initialize(Recog *recog)
{
char *arg = NULL;
ADIn *adin;
Jconf *jconf;
#ifdef ENABLE_PLUGIN
FUNC_INT func;
int sid;
#endif
adin = recog->adin;
jconf = recog->jconf;
jlog("STAT: ###### initialize input device\n");
/* select input device: file, mic, netaudio, etc... */
#ifdef ENABLE_PLUGIN
sid = jconf->input.plugin_source;
if (sid >= 0) {
/* set plugin properties and functions to adin */
func = (FUNC_INT) plugin_get_func(sid, "adin_get_configuration");
if (func == NULL) {
jlog("ERROR: invalid plugin: adin_get_configuration() not exist\n");
return FALSE;
}
adin->silence_cut_default = (*func)(1);
adin->enable_thread = (*func)(2);
adin->ad_standby = (boolean (*)(int, void *)) plugin_get_func(sid, "adin_standby");
adin->ad_begin = (boolean (*)(char *)) plugin_get_func(sid, "adin_open");
adin->ad_end = (boolean (*)()) plugin_get_func(sid, "adin_close");
adin->ad_resume = (boolean (*)()) plugin_get_func(sid, "adin_resume");
adin->ad_pause = (boolean (*)()) plugin_get_func(sid, "adin_pause");
adin->ad_terminate = (boolean (*)()) plugin_get_func(sid, "adin_terminate");
adin->ad_read = (int (*)(SP16 *, int)) plugin_get_func(sid, "adin_read");
adin->ad_input_name = (char * (*)()) plugin_get_func(sid, "adin_input_name");
if (adin->ad_read == NULL) {
jlog("ERROR: m_adin: selected plugin has no function adin_read()\n");
return FALSE;
}
} else {
#endif
/* built-in */
if (adin_select(adin, jconf->input.speech_input, jconf->input.device) == FALSE) {
jlog("ERROR: m_adin: failed to select input device\n");
return FALSE;
}
/* set sampling frequency and device-dependent configuration
(argument is device-dependent) */
switch(jconf->input.speech_input) {
case SP_ADINNET: /* arg: port number */
arg = mymalloc(100);
sprintf(arg, "%d", jconf->input.adinnet_port);
break;
case SP_RAWFILE: /* arg: filename of file list (if any) */
if (jconf->input.inputlist_filename != NULL) {
arg = mymalloc(strlen(jconf->input.inputlist_filename)+1);
strcpy(arg, jconf->input.inputlist_filename);
} else {
arg = NULL;
}
break;
case SP_STDIN:
arg = NULL;
break;
#ifdef USE_NETAUDIO
case SP_NETAUDIO: /* netaudio server/port name */
arg = mymalloc(strlen(jconf->input.netaudio_devname)+1);
strcpy(arg, jconf->input.netaudio_devname);
break;
#endif
}
#ifdef ENABLE_PLUGIN
}
#endif
if (adin_setup_all(adin, jconf, arg) == FALSE) {
return FALSE;
}
if (arg != NULL) free(arg);
return TRUE;
}
/**
* <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;
}
