search_bestfirst_v1.c

julius version 4.12.about sound recognition.

  maxwn = r->lm->winfo->maxwn + 10;
  dwrk = &(r->pass2);

  dwrk->wordtrellis[0] = (LOGPROB *)mymalloc(sizeof(LOGPROB) * maxwn);
  dwrk->wordtrellis[1] = (LOGPROB *)mymalloc(sizeof(LOGPROB) * maxwn);

  dwrk->g = (LOGPROB *)mymalloc(sizeof(LOGPROB) * r->peseqlen);

  dwrk->phmmlen_max = r->lm->winfo->maxwlen + 2;
  dwrk->phmmseq = (HMM_Logical **)mymalloc(sizeof(HMM_Logical *) * dwrk->phmmlen_max);
  if (r->lm->config->enable_iwsp && r->am->hmminfo->multipath) {
    dwrk->has_sp = (boolean *)mymalloc(sizeof(boolean) * dwrk->phmmlen_max);
  } else {
    dwrk->has_sp = NULL;
  }

  dwrk->wend_token_frame[0] = NULL;
  dwrk->wend_token_frame[1] = NULL;
  dwrk->wend_token_gscore[0] = NULL;
  dwrk->wend_token_gscore[1] = NULL;
#ifdef GRAPHOUT_PRECISE_BOUNDARY
  if (r->graphout) {
    dwrk->wend_token_frame[0] = (short *)mymalloc(sizeof(short) * maxwn);
    dwrk->wend_token_frame[1] = (short *)mymalloc(sizeof(short) * maxwn);
    dwrk->wend_token_gscore[0] = (LOGPROB *)mymalloc(sizeof(LOGPROB) * maxwn);
    dwrk->wend_token_gscore[1] = (LOGPROB *)mymalloc(sizeof(LOGPROB) * maxwn);
  }
#endif
  
}

/** 
 * <JA>
 * 1帽胳尸のトレリス纷换脱のワ〖クエアリアを豺庶
 * 
 * </JA>
 * <EN>
 * Free the work area for trellis computation of a word.
 * 
 * </EN>
 * @callgraph
 * @callergraph
 */
void
free_wordtrellis(StackDecode *dwrk)
{
  int i;

  free(dwrk->wordtrellis[0]);
  free(dwrk->wordtrellis[1]);
  free(dwrk->g);
  free(dwrk->phmmseq);
  if (dwrk->has_sp) {
    free(dwrk->has_sp);
    dwrk->has_sp = NULL;
  }
#ifdef GRAPHOUT_PRECISE_BOUNDARY
  for(i=0;i<2;i++) {
    if (dwrk->wend_token_frame[i]) {
      free(dwrk->wend_token_frame[i]);
      dwrk->wend_token_frame[i] = NULL;
    }
    if (dwrk->wend_token_gscore[i]) {
      free(dwrk->wend_token_gscore[i]);
      dwrk->wend_token_gscore[i] = NULL;
    }
  }
#endif
}


/**********************************************************************/
/************ 簿棱の涟羹き锑刨纷换                  *******************/
/************ Compute forward score of a hypothesis *******************/
/**********************************************************************/

/** 
 * <JA>
 * 呵姜觉轮への莲败澄唯の呵络猛を滇める (multipath)
 * 
 * @param tr [in] 莲败乖误
 * @param state_num [in] 觉轮眶
 * 
 * @return 呵姜觉轮への莲败澄唯への呵络猛を手す. 
 * </JA>
 * <EN>
 * Get the maximum transition log probability to final state. (multipath)
 * 
 * @param tr [in] transition matrix
 * @param state_num [in] number of states
 * 
 * @return the maximum log probability of transition to the final state.
 * </EN>
 */
static LOGPROB
get_max_out_arc(HTK_HMM_Trans *tr, int state_num)
{
  LOGPROB max_a;
  int afrom;
  LOGPROB a;
  
  max_a = LOG_ZERO;
  for (afrom = 0; afrom < state_num - 1; afrom++) {
    a = tr->a[afrom][state_num-1];
    if (max_a < a) max_a = a;
  }
  return(max_a);
}

/** 
 * <JA>
 * 不燎の叫蜗觉轮への莲败澄唯の呵络猛を滇める.  (multipath)
 * 
 * @param l [in] 不燎
 * 
 * @return 叫蜗觉轮への莲败澄唯の呵络猛を手す. 
 * </JA>
 * <EN>
 * Get the maximum transition log probability outside a phone. (multipath)
 * 
 * @param l [in] phone
 * 
 * @return the maximum transition log probability outside a phone.
 * </EN>
 */
static LOGPROB
max_out_arc(HMM_Logical *l)
{
  return(get_max_out_arc(hmm_logical_trans(l), hmm_logical_state_num(l)));
}

/** 
 * <JA>
 * 呵稿の1帽胳の涟羹きトレリスを纷换して·矢簿棱の涟羹き锑刨を构糠する. 
 * 
 * @param now [i/o] 矢簿棱
 * @param param [in] 掐蜗パラメ〖タ误
 * @param r [in] 千急借妄インスタンス
 * 
 * </JA>
 * <EN>
 * Compute the forward viterbi for the last word to update forward scores
 * and ready for word connection.
 * 
 * @param now [i/o] hypothesis
 * @param param [in] input parameter vectors
 * @param r [in] recognition process instance
 * 
 * </EN>
 * @callgraph
 * @callergraph
 */
void
scan_word(NODE *now, HTK_Param *param, RecogProcess *r)
{
  int   i,t, j;
  HMM *whmm;
  A_CELL *ac;
  WORD_ID word;
  LOGPROB tmpmax, tmptmp, score1;
  int startt = 0, endt;
  int wordhmmnum;
  LOGPROB tmpmax_store, store_point_maxarc; /* multipath */
  LOGPROB tmpmax2 = LOG_ZERO;
  int phmmlen;
  HMM_Logical *ret, *wend;
  int store_point;
  int crossword_point = 0;
  boolean back_rescan = FALSE;
  boolean node_exist_p;
  int tn;		       ///< Temporal pointer to current buffer
  int tl;		       ///< Temporal pointer to previous buffer

  /* store global values to local for rapid access */
  WORD_INFO *winfo;
  HTK_HMM_INFO *hmminfo;
  LOGPROB *framemaxscore;
  int peseqlen;
  boolean ccd_flag;
  boolean enable_iwsp;
#ifdef SCAN_BEAM
  LOGPROB scan_beam_thres;
#endif
  StackDecode *dwrk;

  winfo = r->lm->winfo;
  hmminfo = r->am->hmminfo;
  dwrk = &(r->pass2);
  peseqlen = r->peseqlen;
  framemaxscore = r->pass2.framemaxscore;
  ccd_flag = r->ccd_flag;
  enable_iwsp = r->lm->config->enable_iwsp; /* multipath */
#ifdef SCAN_BEAM
  scan_beam_thres = r->config->pass2.scan_beam_thres;
#endif

  if (hmminfo->multipath) {
    store_point = -1;
  } else {
    store_point = 0;
  }
  
  /* ----------------------- prepare HMM ----------------------- */

  if (ccd_flag) {
    /* 木涟の不燎があれば·そこまでさかのぼって scan する */
    /* if there are any last phone, enable backscan */
    if (now->last_ph == NULL) {
      /* initial score: now->g[] */
      /* scan range: phones in now->seq[now->seqnum-1] */
      back_rescan = FALSE;
    } else {
      /* initial score: now->g_prev[] (1-phone before)*/
      /* scan range: phones in now->seq[now->seqnum-1] + now->last_ph */
      back_rescan = TRUE;
    }
  }
#ifdef TCD
  if (now->last_ph != NULL) {
    jlog("DEBUG: inherited last_ph: %s\n", (now->last_ph)->name);
    if (now->last_ph_sp_attached) jlog("DEBUG: (sp attached)\n"); /* multipath */
  } else {
    jlog("DEBUG: no last_ph inherited\n");
  }
#endif

  /* scan 认跋尸のHMMを洁洒 */
  /* prepare HMM of the scan range */
  word = now->seq[now->seqnum-1];

  if (ccd_flag) {

    if (back_rescan) {
      
      /* scan range: phones in now->seq[now->seqnum-1] + now->last_ph */
      
      phmmlen = winfo->wlen[word] + 1;
      if (phmmlen > dwrk->phmmlen_max) {
	j_internal_error("scan_word: num of phonemes in a word exceed phmmlenmax (%d) ?\n", dwrk->phmmlen_max);
      }
      for (i=0;i<phmmlen - 2;i++) dwrk->phmmseq[i] = winfo->wseq[word][i];
      if (enable_iwsp && hmminfo->multipath) {
	for (i=0;i<phmmlen - 2;i++) dwrk->has_sp[i] = FALSE;
      }

      /* 呵姜帽胳と last_ph 粗の帽胳粗triphoneを雇胃 */
      /* consider cross-word context dependency between the last word and now->last_ph */
      wend = winfo->wseq[word][winfo->wlen[word]-1];
      ret = get_right_context_HMM(wend, now->last_ph->name, hmminfo);
      if (ret == NULL) {	/* triphone not found */
	/* fallback to the original bi/mono-phone */
	/* error if the original is pseudo phone (not explicitly defined
	   in hmmdefs/hmmlist) */
	/* exception: word with 1 phone (triphone may exist in the next expansion */
	if (winfo->wlen[word] > 1 && wend->is_pseudo) {
	  error_missing_right_triphone(wend, now->last_ph->name);
	}
	dwrk->phmmseq[phmmlen-2] = wend;
      } else {
	dwrk->phmmseq[phmmlen-2] = ret;
      }
      ret = get_left_context_HMM(now->last_ph, wend->name, hmminfo);
      if (ret == NULL) {
	/* fallback to the original bi/mono-phone */
	/* error if the original is pseudo phone (not explicitly defined
	   in hmmdefs/hmmlist) */
	if (now->last_ph->is_pseudo) {
	  error_missing_left_triphone(now->last_ph, wend->name);
	}
	dwrk->phmmseq[phmmlen-1] = now->last_ph;
      } else {
	dwrk->phmmseq[phmmlen-1] = ret;
      }

      if (enable_iwsp && hmminfo->multipath) {
	dwrk->has_sp[phmmlen-2] = TRUE;
	dwrk->has_sp[phmmlen-1] = now->last_ph_sp_attached;
      }
 
#ifdef TCD
      jlog("DEBUG: w=");
      for(i=0;i<winfo->wlen[word];i++) {
	jlog(" %s",(winfo->wseq[word][i])->name);
	if (enable_iwsp && hmminfo->multipath && dwrk->has_sp[i]) jlog("(sp)");
      }
      jlog(" | %s\n", (now->last_ph)->name);
      if (hmminfo->multipath && now->last_ph_sp_attached) jlog("DEBUG:   (sp)\n");
      jlog("DEBUG: scan for:");
      
      for (i=0;i<phmmlen;i++) {
	jlog(" %s", dwrk->phmmseq[i]->name);
 	if (enable_iwsp && hmminfo->multipath && dwrk->has_sp[i]) jlog("(sp)");
      }
      jlog("\n");
#endif

      /* 帽胳HMMを侯る */
      /* make word HMM */
      whmm = new_make_word_hmm(hmminfo, dwrk->phmmseq, phmmlen, (enable_iwsp && hmminfo->multipath) ? dwrk->has_sp : NULL);
      if (whmm == NULL) {
	j_internal_error("Error: failed to make word hmm for word #%d \"%s [%s]\"\n", word, winfo->wname[word], winfo->woutput[word]);
      }
      
      /* backscan なので·纷换涟の g[] 介袋猛は now->g_prev[] を蝗脱 */
      /* As backscan enabled, the initial forward score g[] is set by
	 now->g_prev[] */
      for (t=0;t<peseqlen;t++) {
	dwrk->g[t]=now->g_prev[t];

      }
      
      /* 肌檬脱のg_prevを呈羌するノ〖ド疤弥を肋年 */
      /* set where to store scores as new g_prev[] for the next backscan
	 in the HMM */
      if (hmminfo->multipath) {
	store_point = hmm_logical_state_num(dwrk->phmmseq[0]) - 2;
	store_point_maxarc = max_out_arc(dwrk->phmmseq[0]);
	if (enable_iwsp && dwrk->has_sp[0]) {
	  store_point += hmm_logical_state_num(hmminfo->sp) - 2;
	  if (store_point_maxarc < max_out_arc(hmminfo->sp)) {
	    store_point_maxarc = max_out_arc(hmminfo->sp);
	  }
	}
      } else {
	store_point = hmm_logical_state_num(dwrk->phmmseq[0]) - 2 - 1;
      }
      /* scan面に木涟帽胳とこの帽胳をまたぐ眷疥を肋年 */
      /* set where is the connection point of the last word in the HMM */
      if (hmminfo->multipath) {
	crossword_point = whmm->len - hmm_logical_state_num(dwrk->phmmseq[phmmlen-1]);
	if (enable_iwsp && dwrk->has_sp[phmmlen-1]) {
	  crossword_point -= hmm_logical_state_num(hmminfo->sp) - 2;
	}
      } else {
	crossword_point = whmm->len - (hmm_logical_state_num(dwrk->phmmseq[phmmlen-1]) - 2) - 1;
      }
      
    } else {			/* not backscan mode */
      
      /* scan range: phones in now->seq[now->seqnum-1] */
      
#ifdef TCD
      jlog("DEBUG: scan(org):");
      for (i=0;i<winfo->wlen[word];i++) {
	jlog(" %s", (winfo->wseq[word][i])->name);
      }
      jlog("\n");
#endif

      if (enable_iwsp && hmminfo->multipath) {
	/* 涩妥ならばショ〖トポ〖ズを洞み哈む疤弥を回年する */
	for(i=0;i<winfo->wlen[word];i++) {
	  dwrk->has_sp[i] = FALSE;
	}
	dwrk->has_sp[winfo->wlen[word]-1] = TRUE;
      }
      
      /* 帽胳HMMを侯る */
      /* make word HMM */
      whmm = new_make_word_hmm(hmminfo, winfo->wseq[word], winfo->wlen[word], (enable_iwsp && hmminfo->multipath) ? dwrk->has_sp : NULL);
      if (whmm == NULL) {
	j_internal_error("Error: failed to make word hmm for word #%d \"%s [%s]\"\n", word, winfo->wname[word], winfo->woutput[word]);
      }
      
      /* 纷换涟の g[] 介袋猛は now->g[] を蝗脱 */
      /* the initial forward score g[] is set by now->g[] */
      for (t=0;t<peseqlen;t++) {
	dwrk->g[t]=now->g[t];
      }
      
      /* 肌檬脱のg_prevを呈羌するノ〖ド疤弥を肋年 */
      /* set where to store scores as new g_prev[] for the next backscan
	 in the HMM */
      if (hmminfo->multipath) {
	store_point = hmm_logical_state_num(winfo->wseq[word][0]) - 2;
	store_point_maxarc = max_out_arc(winfo->wseq[word][0]);
	if (enable_iwsp && dwrk->has_sp[0]) {
	  store_point += hmm_logical_state_num(hmminfo->sp) - 2;
	  if (store_point_maxarc < max_out_arc(hmminfo->sp)) {
	    store_point_maxarc = max_out_arc(hmminfo->sp);
	  }
	}
      } else {
	store_point = hmm_logical_state_num(winfo->wseq[word][0]) - 2 - 1;
      }

      /* scan面に木涟帽胳とこの帽胳をまたぐ眷疥は·なし */
      /* the connection point of the last word is not exist in the HMM */
      crossword_point = -1;
    }
    
  } else {			/* ccd_flag == FALSE */

    if (enable_iwsp && hmminfo->multipath) {
      /* 涩妥ならばショ〖トポ〖ズを洞み哈む疤弥を回年する */
      for(i=0;i<winfo->wlen[word];i++) {
	dwrk->has_sp[i] = FALSE;
      }
      dwrk->has_sp[winfo->wlen[word]-1] = TRUE;
    }

    /* 不燎茨董润巴赂の眷圭は帽姐に呵姜帽胳尸の HMM を侯喇 */
    /* for monophone: simple make HMM for the last word */
    whmm = new_make_word_hmm(hmminfo, winfo->wseq[word], winfo->wlen[word], (enable_iwsp && hmminfo->multipath) ? dwrk->has_sp : NULL);
    if (whmm == NULL) {
      j_internal_error("Error: failed to make word hmm for word #%d \"%s [%s]\"\n", word, winfo->wname[word], winfo->woutput[word]);
    }

    /* 纷换涟の g[] 介袋猛は now->g[] を蝗脱 */
    /* the initial forward score g[] is set by now->g[] */
    for (t=0;t<peseqlen;t++) {
      dwrk->g[t]=now->g[t];
    }
    
  }

#ifdef TCD
  jlog("DEBUG: whmm len	  = %d\n",whmm->len);
  jlog("DEBUG: crossword_point = %d\n", crossword_point);
  jlog("DEBUG: g[] store point = %d\n", store_point);
#endif

  wordhmmnum = whmm->len;
  if (wordhmmnum >= winfo->maxwn + 10) {
    j_internal_error("scan_word: word too long (>%d)\n", winfo->maxwn + 10);
  }

#ifndef GRAPHOUT_PRECISE_BOUNDARY
  if (r->graphout) {