search_bestfirst_v2.c

julius version 4.12.about sound recognition.

 */
void
next_word(NODE *now, NODE *new, NEXTWORD *nword, HTK_Param *param, RecogProcess *r)
{
  HMM_Logical *lastphone, *newphone;
  LOGPROB *g_src;
  int   t;
  int lastword;
  int   i;
  LOGPROB a_value;
  LOGPROB tmpp;
  int   startt;
  int word;
  TRELLIS_ATOM *tre;
  LOGPROB totalscore;
  BACKTRELLIS *backtrellis;
  WORD_INFO *winfo;
  HTK_HMM_INFO *hmminfo;
  int peseqlen;
  boolean ccd_flag;
  StackDecode *dwrk;

  dwrk = &(r->pass2);
  backtrellis = r->backtrellis;
  winfo = r->lm->winfo;
  hmminfo = r->am->hmminfo;
  peseqlen = r->peseqlen;
  ccd_flag = r->ccd_flag;

  word = nword->id;
  lastword = now->seq[now->seqnum-1];

  /* lastphone (木涟帽胳の黎片不燎) を洁洒 */
  /* prepare lastphone (head phone of previous word) */
  if (ccd_flag) {
    /* 呵姜不燎 triphone を儡鲁帽胳に柴わせて恃步 */
    /* modify triphone of last phone according to the next word */
    lastphone = get_left_context_HMM(now->last_ph, winfo->wseq[word][winfo->wlen[word]-1]->name, hmminfo);
    if (lastphone == NULL) {
      /* 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 (now->last_ph->is_pseudo){
	error_missing_left_triphone(now->last_ph, winfo->wseq[word][winfo->wlen[word]-1]->name);
      }
      lastphone = now->last_ph;
    }
  }

  /* newphone (儡鲁帽胳の琐萨不燎) を洁洒 */
  /* prepare newphone (tail phone of next word) */
  if (ccd_flag) {
    newphone = get_right_context_HMM(winfo->wseq[word][winfo->wlen[word]-1], now->last_ph->name, hmminfo);
    if (newphone == NULL) {
      /* 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 && winfo->wseq[word][winfo->wlen[word]-1]->is_pseudo){
	error_missing_right_triphone(winfo->wseq[word][winfo->wlen[word]-1], now->last_ph->name);
      }
      newphone = winfo->wseq[word][winfo->wlen[word]-1];
    }
  } else {
    newphone = winfo->wseq[word][winfo->wlen[word]-1];
  }
  
  /* 帽胳事び、DFA觉轮戎规、咐胳スコアを new へ费镜ˇ构糠 */
  /* inherit and update word sequence, DFA state and total LM score to 'new' */
  new->score = LOG_ZERO;
  for (i=0;i< now->seqnum;i++){
    new->seq[i] = now->seq[i];
#ifdef CM_SEARCH
#ifdef CM_MULTIPLE_ALPHA
    memcpy(new->cmscore[i], now->cmscore[i], sizeof(LOGPROB) * r->config->annotate.cm_alpha_num);
#else
    new->cmscore[i] = now->cmscore[i];
#endif
#endif /* CM_SEARCH */
  }
  new->seq[i] = word;
  new->seqnum = now->seqnum+1;
  new->state = nword->next_state;
  new->totallscore = now->totallscore + nword->lscore;
  if (ccd_flag) {
    /* 肌簿棱の旺悟攫鼠として瘦赂 */
    /* keep the lastphone for next scan_word() */
    new->last_ph = lastphone;
    new->last_ph_sp_attached = now->last_ph_sp_attached;
  }

  if (ccd_flag) {
    /* 呵稿の1不燎(lastphone)尸をscanし·构糠したスコアを new に瘦赂 */
    /* scan the lastphone and set the updated score to new->g[] */
    do_viterbi_next_word(now, new, lastphone,
			 hmminfo->multipath ? now->last_ph_sp_attached : FALSE,
			 param, r);
    g_src = new->g;
  } else {
    g_src = now->g;
#ifdef GRAPHOUT_PRECISE_BOUNDARY
    if (r->graphout) {
      memcpy(new->wordend_frame, now->wordend_frame, sizeof(short)*peseqlen);
      memcpy(new->wordend_gscore, now->wordend_gscore, sizeof(LOGPROB)*peseqlen);
    }
#endif
  }
      
  /* 肌搀の scan_word に洒えて new->g[] を恃构しておく */
  /* prepare new->g[] for next scan_word() */
  if (hmminfo->multipath) {
    startt = peseqlen-1;
  } else {
    startt = peseqlen-2;
  }
  i = hmm_logical_state_num(newphone);
  a_value = (hmm_logical_trans(newphone))->a[i-2][i-1];
  if (hmminfo->multipath) {
    for(t=0; t <= startt; t++) {
      new->g[t] = g_src[t] + nword->lscore;
    }
  } else {
    for(t=0; t <= startt; t++) {
      new->g[t] = g_src[t+1] + a_value + nword->lscore;
    }
  }

  /***************************************************************************/
  /* 涟羹き(妈２パス),稿ろ羹き(妈１パス)トレリスを儡鲁し呵锑儡鲁爬を斧つける */
  /* connect forward/backward trellis to look for the best connection time   */
  /***************************************************************************/
  /*-----------------------------------------------------------------*/
  /* 帽胳トレリスを玫して, 肌帽胳の呵锑儡鲁爬を券斧する */
  /* determine the best connection time of the new word, seeking the word
     trellis */
  /*-----------------------------------------------------------------*/

  if (r->lmtype == LM_DFA && !r->config->pass2.looktrellis_flag) {
    /* すべてのフレ〖ムにわたって呵锑を玫す */
    /* search for best trellis word throughout all frame */
    for(t = startt; t >= 0; t--) {
      tre = bt_binsearch_atom(backtrellis, t, (WORD_ID) word);
      if (tre == NULL) continue;
      totalscore = new->g[t] + tre->backscore;
      if (! hmminfo->multipath) {
	if (newphone->is_pseudo) {
	  tmpp = outprob_cd(&(r->am->hmmwrk), t, &(newphone->body.pseudo->stateset[newphone->body.pseudo->state_num-2]), param);
	} else {
	  tmpp = outprob_state(&(r->am->hmmwrk), t, newphone->body.defined->s[newphone->body.defined->state_num-2], param);
	}
	totalscore += tmpp;
      }
      if (new->score < totalscore) {
	new->score = totalscore;
	new->bestt = t;
	new->estimated_next_t = tre->begintime - 1;
	new->tre = tre;
      }
    }

    return;
  }

  /* 呵稿に徊救したTRELLIS_ATOMの姜眉箕粗の涟稿 */
  /* newの夸年箕粗は·惧淡で何脱したTRELLIS_ATOMの幌眉箕粗 */

  /* この鸥倡帽胳のトレリス惧の姜眉箕粗の涟稿のみスキャンする
     涟稿に息鲁して赂哼するフレ〖ムについてのみ纷换 */
  /* search for best trellis word only around the estimated time */
  /* 1. search forward */
  for(t = (nword->tre)->endtime; t >= 0; t--) {
    tre = bt_binsearch_atom(backtrellis, t, (WORD_ID) word);
    if (tre == NULL) break;	/* go to 2 if the trellis word disappear */
    totalscore = new->g[t] + tre->backscore;
    if (! hmminfo->multipath) {
      if (newphone->is_pseudo) {
	tmpp = outprob_cd(&(r->am->hmmwrk), t, &(newphone->body.pseudo->stateset[newphone->body.pseudo->state_num-2]), param);
      } else {
	tmpp = outprob_state(&(r->am->hmmwrk), t, newphone->body.defined->s[newphone->body.defined->state_num-2], param);
      }
      totalscore += tmpp;
    }
    if (new->score < totalscore) {
      new->score = totalscore;
      new->bestt = t;
      new->estimated_next_t = tre->begintime - 1;
      new->tre = tre;
    }
  }
  /* 2. search bckward */
  for(t = (nword->tre)->endtime + 1; t <= startt; t++) {
    tre = bt_binsearch_atom(backtrellis, t, (WORD_ID) word);
    if (tre == NULL) break;	/* end if the trellis word disapper */
    totalscore = new->g[t] + tre->backscore;
    if (! hmminfo->multipath) {
      if (newphone->is_pseudo) {
	tmpp = outprob_cd(&(r->am->hmmwrk), t, &(newphone->body.pseudo->stateset[newphone->body.pseudo->state_num-2]), param);
      } else {
	tmpp = outprob_state(&(r->am->hmmwrk), t, newphone->body.defined->s[newphone->body.defined->state_num-2], param);
      }
      totalscore += tmpp;
    }
    if (new->score < totalscore) {
      new->score = totalscore;
      new->bestt = t;
      new->estimated_next_t = tre->begintime - 1;
      new->tre = tre;
    }
  }

  /* set current LM score */
  new->lscore = nword->lscore;
  
}


/**********************************************************************/
/********** 介袋簿棱の栏喇                 ****************************/
/********** Generate an initial hypothesis ****************************/
/**********************************************************************/

/** 
 * <JA>
 * 涂えられた帽胳から介袋簿棱を栏喇する. 
 * 
 * @param new [out] 糠たに栏喇された簿棱が呈羌される
 * @param nword [in] 介袋簿棱帽胳の攫鼠
 * @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>
 * Generate an initial hypothesis from given word.
 * 
 * @param new [out] pointer to save the newly generated hypothesis
 * @param nword [in] words of the first candidates
 * @param param [in] input parameter vector
 * @param r [in] recognition process instance
 *
 * </EN>
 * @callgraph
 * @callergraph
 */
void
start_word(NODE *new, NEXTWORD *nword, HTK_Param *param, RecogProcess *r)
{
  HMM_Logical *newphone;
  WORD_ID word;
  TRELLIS_ATOM *tre = NULL;
  LOGPROB tmpp;
  int t;

  BACKTRELLIS *backtrellis;
  WORD_INFO *winfo;

  int peseqlen;
  boolean ccd_flag;
  boolean multipath;

  backtrellis = r->backtrellis;
  winfo = r->lm->winfo;
  peseqlen = r->peseqlen;
  ccd_flag = r->ccd_flag;
  multipath = r->am->hmminfo->multipath;

  /* initialize data */
  word = nword->id;
  new->score = LOG_ZERO;
  new->seqnum = 1;
  new->seq[0] = word;

  new->state = nword->next_state;
  new->totallscore = nword->lscore;

  /* set current LM score */
  new->lscore = nword->lscore;

  /* cross-word triphone need not be handled on startup */
  newphone = winfo->wseq[word][winfo->wlen[word]-1];
  if (ccd_flag) {
    new->last_ph = NULL;
  }
  
  if (r->lmtype == LM_PROB) {
    new->g[peseqlen-1] = nword->lscore;
  } else {
    new->g[peseqlen-1] = 0;
  }
  
  for (t=peseqlen-1; t>=0; t--) {
    tre = bt_binsearch_atom(backtrellis, t, word);
    if (tre != NULL) {
      if (r->graphout) {
	new->bestt = peseqlen-1;
      } else {
	new->bestt = t;
      }
      new->score = new->g[peseqlen-1] + tre->backscore;
      if (! multipath) {
	if (newphone->is_pseudo) {
	  tmpp = outprob_cd(&(r->am->hmmwrk), peseqlen-1, &(newphone->body.pseudo->stateset[newphone->body.pseudo->state_num-2]), param);
	} else {
	  tmpp = outprob_state(&(r->am->hmmwrk), peseqlen-1, newphone->body.defined->s[newphone->body.defined->state_num-2], param);
	}
	new->score += tmpp;
      }
      new->estimated_next_t = tre->begintime - 1;
      new->tre = tre;
      break;
    }
  }
  if (tre == NULL) {		/* no word in backtrellis */
    new->score = LOG_ZERO;
  }
}

/** 
 * <JA>
 * 姜眉借妄¨姜眉まで茫した矢簿棱の呵姜弄なスコアをセットする. 
 * 
 * @param now [in] 姜眉まで茫した簿棱
 * @param new [out] 呵姜弄な矢簿棱のスコアを呈羌する眷疥へのポインタ
 * @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>
 * Hypothesis termination: set the final sentence scores of hypothesis
 * that has already reached to the end.
 * 
 * @param now [in] hypothesis that has already reached to the end
 * @param new [out] pointer to save the final sentence information
 * @param param [in] input parameter vectors
 * @param r [in] recognition process instance
 * </EN>
 * @callgraph
 * @callergraph
 */
void
last_next_word(NODE *now, NODE *new, HTK_Param *param, RecogProcess *r)
{
  cpy_node(new, now);
  if (r->ccd_flag) {
    /* 呵姜不燎尸を viterbi して呵姜スコアを肋年 */
    /* scan the last phone and update the final score */
    if (r->am->hmminfo->multipath) {
      do_viterbi_next_word(now, new, now->last_ph, now->last_ph_sp_attached, param, r);
      new->score = new->final_g;
    } else {
      do_viterbi_next_word(now, new, now->last_ph, FALSE, param, r);
      new->score = new->g[0];
    }
  } else {
    if (r->am->hmminfo->multipath) {
      new->score = now->final_g;
    } else {
      new->score = now->g[0];
    }
#ifdef GRAPHOUT_PRECISE_BOUNDARY
    if (r->graphout) {
      /* last boundary has moved to [peseqlen-1] in last scan_word() */
      memcpy(new->wordend_frame, now->wordend_frame, sizeof(short)*r->peseqlen);
      memcpy(new->wordend_gscore, now->wordend_gscore, sizeof(LOGPROB)*r->peseqlen);
    }
#endif
  }
}

#endif /* PASS2_STRICT_IWCD */

/* end of file */