search_bestfirst_v2.c

julius version 4.12.about sound recognition.

  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->wef = NULL;
  dwrk->wes = 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->wef = (short *)mymalloc(sizeof(short) * r->peseqlen);
    dwrk->wes = (LOGPROB *)mymalloc(sizeof(LOGPROB) * r->peseqlen);
    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)
{
  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
  if (dwrk->wef) {
    free(dwrk->wef);
    free(dwrk->wes);
    free(dwrk->wend_token_frame[0]);
    free(dwrk->wend_token_frame[1]);
    free(dwrk->wend_token_gscore[0]);
    free(dwrk->wend_token_gscore[1]);
    dwrk->wef = NULL;
  }
#endif
}


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

/* 涂えられた不燎のならび phmmseq[0..phmmlen-1]に滦してviterbi纷换を乖う. 
   g[0..framelen-1] のスコアを介袋猛として g_new[0..framelen-1]に构糠猛を洛掐. 
   呵你 least_frame まではscanする. */
/* Viterbi computation for the given phoneme sequence 'phmmseq[0..phmmlen-1]'
   with g[0..framelen-1] as initial values.  The results are stored in
   g_new[0..framelen-1].  Scan should not terminate at least it reaches
   'least_frame'. */
/** 
 * <JA>
 * 涂えられた不燎の事びに滦して Viterbi 纷换を乖い·涟羹きスコアを
 * 构糠する绕脱簇眶. 
 * 
 * @param g [in] 附哼の箕粗ごとの涟羹きスコア
 * @param g_new [out] 构糠稿の糠たな涟羹きスコアを呈羌するバッファ
 * @param phmmseq [in] 不燎HMMの事び
 * @param has_sp [in] short-pause location
 * @param phmmlen [in] @a phmmseq の墓さ
 * @param param [in] 掐蜗パラメ〖タ
 * @param framelen [in] 掐蜗フレ〖ム墓
 * @param least_frame [in] ビ〖ム肋年箕·このフレ〖ム眶笆惧は Viterbi纷换する
 * @param final_g [in] final g scores
 * @param wordend_frame_src [in] 附哼の帽胳姜眉フレ〖ムト〖クン
 * @param wordend_frame_dst [out] 构糠稿の糠たな帽胳姜眉フレ〖ムト〖クン
 * @param wordend_gscore_src [in] 附哼の帽胳姜眉スコアト〖クン
 * @param wordend_gscore_dst [out] 构糠稿の糠たな帽胳姜眉スコアト〖クン
 * @param r [in] recognition process instance
 * </JA>
 * <EN>
 * Generic function to perform Viterbi path updates for given phoneme
 * sequence.
 * 
 * @param g [in] current forward scores at each input frame
 * @param g_new [out] buffer to save the resulting score updates
 * @param phmmseq [in] phoneme sequence to perform Viterbi
 * @param has_sp [in] short-pause location
 * @param phmmlen [in] length of @a phmmseq.
 * @param param [in] input parameter vector
 * @param framelen [in] input frame length to compute
 * @param least_frame [in] Least frame length to force viterbi even with beam
 * @param final_g [in] final g scores
 * @param wordend_frame_src [in] current word-end frame tokens
 * @param wordend_frame_dst [out] buffer to store updated word-end frame tokens
 * @param wordend_gscore_src [in] current word-end score tokens
 * @param wordend_gscore_dst [out] buffer to store updated word-end score tokens
 * @param r [in] recognition process instance
 * 
 * </EN>
 */
static void
do_viterbi(LOGPROB *g, LOGPROB *g_new, HMM_Logical **phmmseq, boolean *has_sp, int phmmlen, HTK_Param *param, int framelen, int least_frame, LOGPROB *final_g, short *wordend_frame_src, short *wordend_frame_dst, LOGPROB *wordend_gscore_src, LOGPROB *wordend_gscore_dst, RecogProcess *r) /* has_sp and final_g is for multipath only */
{
  HMM *whmm;			/* HMM */
  int wordhmmnum;		/* length of above */
  int startt;			/* scan start frame */
  LOGPROB tmpmax,tmpscore;	/* variables for Viterbi process */
  A_CELL *ac;
  int t,i,j;
  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 */
  StackDecode *dwrk;
  WORD_INFO *winfo;
  HTK_HMM_INFO *hmminfo;
  LOGPROB *framemaxscore;
#ifdef SCAN_BEAM
  LOGPROB scan_beam_thres;
#endif

  dwrk = &(r->pass2);
  winfo = r->lm->winfo;
  hmminfo = r->am->hmminfo;
  framemaxscore = r->pass2.framemaxscore;
#ifdef SCAN_BEAM
  scan_beam_thres = r->config->pass2.scan_beam_thres;
#endif


#ifdef TCD
  jlog("DEBUG: scan for:");
  for (i=0;i<phmmlen;i++) {
    jlog(" %s", phmmseq[i]->name);
  }
  jlog("\n");
#endif
  
  /* 帽胳HMMを侯る */
  /* make word HMM */
  whmm = new_make_word_hmm(hmminfo, phmmseq, phmmlen, has_sp);
  if (whmm == NULL) {
    j_internal_error("Error: failed to make word hmm\n");
  }
  wordhmmnum = whmm->len;
  if (wordhmmnum >= winfo->maxwn + 10) {
    j_internal_error("do_viterbi: word too long (>%d)\n", winfo->maxwn + 10);
  }

  /* scan倡幌爬を浮瑚 -> starttへ*/
  /* search for the start frame -> set to startt */
  for(t = framelen-1; t >=0 ; t--) {
    if (
#ifdef SCAN_BEAM
	g[t] > framemaxscore[t] - scan_beam_thres &&
#endif
	g[t] > LOG_ZERO) {
      break;
    }
  }
  if (t < 0) {			/* no node has score > LOG_ZERO */
    /* reset all scores and end */
    for(t=0;t<framelen;t++) {
      g_new[t] = LOG_ZERO;
#ifdef GRAPHOUT_PRECISE_BOUNDARY
      if (r->graphout) {
	wordend_frame_dst[t] = -1;
	wordend_gscore_dst[t] = LOG_ZERO;
      }
#endif
    }
    free_hmm(whmm);
    return;
  }
  startt = t;
  
  /* 倡幌爬笆惯[startt+1..framelen-1] の g_new[] をリセット */
  /* clear g_new[] for [startt+1..framelen-1] */
  for(t=framelen-1;t>startt;t--) {
    g_new[t] = LOG_ZERO;
#ifdef GRAPHOUT_PRECISE_BOUNDARY
    if (r->graphout) {
      wordend_frame_dst[t] = -1;
      wordend_gscore_dst[t] = LOG_ZERO;
    }
#endif
  }

  /*****************/
  /* viterbi start */
  /*****************/

  /* set initial swap buffer */
  tn = 0; tl = 1;

#ifdef GRAPHOUT_PRECISE_BOUNDARY
  if (r->graphout) {
    for(i=0;i<wordhmmnum;i++) {
      dwrk->wend_token_frame[tn][i] = -1;
      dwrk->wend_token_gscore[tn][i] = LOG_ZERO;
    }
  }
#endif

  if (! hmminfo->multipath) {
    /* 箕粗 [startt] 惧の猛を介袋步 */
    /* initialize scores on frame [startt] */
    for(i=0;i<wordhmmnum-1;i++) dwrk->wordtrellis[tn][i] = LOG_ZERO;
    dwrk->wordtrellis[tn][wordhmmnum-1] = g[startt] + outprob(&(r->am->hmmwrk), startt, &(whmm->state[wordhmmnum-1]), param);
    g_new[startt] = dwrk->wordtrellis[tn][0];
#ifdef GRAPHOUT_PRECISE_BOUNDARY
    if (r->graphout) {
      dwrk->wend_token_frame[tn][wordhmmnum-1] = wordend_frame_src[startt];
      dwrk->wend_token_gscore[tn][wordhmmnum-1] = wordend_gscore_src[startt];
      wordend_frame_dst[startt] = dwrk->wend_token_frame[tn][0];
      wordend_gscore_dst[startt] = dwrk->wend_token_gscore[tn][0];
    }
#endif
  }
  
  /* メインル〖プ: startt から幌まり 0 に羹かって Viterbi 纷换 */
  /* main loop: start from [startt], and compute Viterbi toward [0] */
  for(t = hmminfo->multipath ? startt : startt - 1; t >= 0; t--) {
    
    /* wordtrellisのワ〖クエリアをスワップ */
    /* swap workarea of wordtrellis */
    i = tn; tn = tl; tl = i;

    node_exist_p = FALSE;	/* TRUE if there is at least 1 survived node in this frame */

    if (! hmminfo->multipath) {
    
      /* 眉のノ〖ド [t][wordhmmnum-1]は·柒婶莲败 か g[]の光い数になる */
      /* the edge node [t][wordhmmnum-1] is either internal transitin or g[] */
      tmpscore = LOG_ZERO;
      for (ac=whmm->state[wordhmmnum-1].ac;ac;ac=ac->next) {
	if (tmpscore < dwrk->wordtrellis[tl][ac->arc] + ac->a) {
	  j = ac->arc;
	  tmpscore = dwrk->wordtrellis[tl][ac->arc] + ac->a;
	}
      }
      if (g[t] > tmpscore) {
	tmpmax = g[t];
#ifdef GRAPHOUT_PRECISE_BOUNDARY
	if (r->graphout) {
	  dwrk->wend_token_frame[tn][wordhmmnum-1] = wordend_frame_src[t];
	  dwrk->wend_token_gscore[tn][wordhmmnum-1] = wordend_gscore_src[t];
	}
#endif
      } else {
	tmpmax = tmpscore;
#ifdef GRAPHOUT_PRECISE_BOUNDARY
	if (r->graphout) {
	  dwrk->wend_token_frame[tn][wordhmmnum-1] = dwrk->wend_token_frame[tl][j];
	  dwrk->wend_token_gscore[tn][wordhmmnum-1] = dwrk->wend_token_gscore[tl][j];
	}
#endif
      }
      
      /* 眉のノ〖ドのスコアエンベロ〖プチェック: 办年升嘲なら皖とす */
      /* check if the edge node is within score envelope */
      if (
#ifdef SCAN_BEAM
	  tmpmax <= framemaxscore[t] - scan_beam_thres ||
#endif
	  tmpmax <= LOG_ZERO
	  ) {
	dwrk->wordtrellis[tn][wordhmmnum-1] = LOG_ZERO;
#ifdef GRAPHOUT_PRECISE_BOUNDARY
	if (r->graphout) {
	  dwrk->wend_token_frame[tn][wordhmmnum-1] = -1;
	  dwrk->wend_token_gscore[tn][wordhmmnum-1] = LOG_ZERO;
	}
#endif
      } else {
	node_exist_p = TRUE;
	dwrk->wordtrellis[tn][wordhmmnum-1] = tmpmax + outprob(&(r->am->hmmwrk), t, &(whmm->state[wordhmmnum-1]), param);
      }

    }

    /* node[wordhmmnum-2..0]についてトレリスを鸥倡 */
    /* expand trellis for node [t][wordhmmnum-2..0] */
    for(i=wordhmmnum-2;i>=0;i--) {
      
      /* 呵锑パスと呵锑スコア tmpmax を斧つける */
      /* find most likely path and the max score 'tmpmax' */
      tmpmax = LOG_ZERO;
      for (ac=whmm->state[i].ac;ac;ac=ac->next) {
	if (hmminfo->multipath) {
	  if (ac->arc == wordhmmnum-1) tmpscore = g[t];
	  else if (t + 1 > startt) tmpscore = LOG_ZERO;
	  else tmpscore = dwrk->wordtrellis[tl][ac->arc];
	  tmpscore += ac->a;
	} else {
	  tmpscore = dwrk->wordtrellis[tl][ac->arc] + ac->a;
	}
	if (tmpmax < tmpscore) {
	  tmpmax = tmpscore;
	  j = ac->arc;
	}
      }
      
      /* スコアエンベロ〖プチェック: 办年升嘲なら皖とす */
      /* check if score of this node is within the score envelope */
      if (
#ifdef SCAN_BEAM
	  tmpmax <= framemaxscore[t] - scan_beam_thres ||
#endif
	  tmpmax <= LOG_ZERO
	  ) {
	/* invalid node */
	dwrk->wordtrellis[tn][i] = LOG_ZERO;
#ifdef GRAPHOUT_PRECISE_BOUNDARY
	if (r->graphout) {
	  dwrk->wend_token_frame[tn][i] = -1;
	  dwrk->wend_token_gscore[tn][i] = LOG_ZERO;
	}
#endif
      } else {
	/* survived node */
	node_exist_p = TRUE;
 	dwrk->wordtrellis[tn][i] = tmpmax;
	if (! hmminfo->multipath || i > 0) {
	  dwrk->wordtrellis[tn][i] += outprob(&(r->am->hmmwrk), t, &(whmm->state[i]), param);
	}
#ifdef GRAPHOUT_PRECISE_BOUNDARY
	if (r->graphout) {
	  if (hmminfo->multipath) {
	    if (j == wordhmmnum-1) {
	      dwrk->wend_token_frame[tn][i] = wordend_frame_src[t];
	      dwrk->wend_token_gscore[tn][i] = wordend_gscore_src[t];
	    } else {
	      dwrk->wend_token_frame[tn][i] = dwrk->wend_token_frame[tl][j];
	      dwrk->wend_token_gscore[tn][i] = dwrk->wend_token_gscore[tl][j];
	    }
	  } else {
	    dwrk->wend_token_frame[tn][i] = dwrk->wend_token_frame[tl][j];
	    dwrk->wend_token_gscore[tn][i] = dwrk->wend_token_gscore[tl][j];
	  }
	}
#endif
      }
    } /* end of node loop */

    /* 箕粗 t のViterbi纷换姜位. 糠たな涟羹きスコア g_new[t] をセット */
    /* Viterbi end for frame [t].  set the new forward score g_new[t] */
    g_new[t] = dwrk->wordtrellis[tn][0];
#ifdef GRAPHOUT_PRECISE_BOUNDARY
    if (r->graphout) {
    /* new wordend */