beam.c

julius version 4.12.about sound recognition.

#endif
	if (r->config->successive.enabled) {
	  /* short-pause segmentation mode */
	  /* 呵姜フレ〖ムに荒った呵络スコアの帽胳 */
	  /* it should be the best trellis word on the last frame */
	  if (maxscore < tmp->backscore) {
	    maxscore = tmp->backscore;
	    best = tmp;
	  }
	} else {
	  /* not segmentation mode */
	  /* 呵姜帽胳は winfo->tail_silwid に盖年 */
	  /* it is fixed to the tail silence model (winfo->tail_silwid) */
	  if (tmp->wid == winfo->tail_silwid && maxscore < tmp->backscore) {
	    maxscore = tmp->backscore;
	    best = tmp;
	    break;
	  }
	}
      }
      if (maxscore != LOG_ZERO) break;
    }

    if (last_time < 0) {		/* not found */
      jlog("WARNING: %02d %s: no tail silence word survived on the last frame, search failed\n", r->config->id, r->config->name);
      r->result.status = J_RESULT_STATUS_FAIL;
      //callback_exec(CALLBACK_RESULT, r);
      return;
    }
  
  }

  if (r->lmtype == LM_DFA) {

    for (last_time = framelen - 1; last_time >= 0; last_time--) {

      /* 琐萨に荒った帽胳の面で呵络スコアの帽胳(cp_endは蝗脱しない) */
      /* the best trellis word on the last frame (not use cp_end[]) */
      maxscore = LOG_ZERO;
      for (i=0;i<backtrellis->num[last_time];i++) {
	tmp = backtrellis->rw[last_time][i];
#ifdef WORD_GRAPH
	/* treat only words on a graph path */
	if (!tmp->within_context) continue;
#endif
	/*      if (dfa->cp_end[winfo->wton[tmp->wid]] == TRUE) {*/
	if (maxscore < tmp->backscore) {
	  maxscore = tmp->backscore;
	  best = tmp;
	}
	/*      }*/
      }
      if (maxscore != LOG_ZERO) break;
    }

    if (last_time < 0) {		/* not found */
      jlog("WARNING: %02d %s: no sentence-end word survived on last beam\n", r->config->id, r->config->name);
      r->result.status = J_RESULT_STATUS_FAIL;
      //callback_exec(CALLBACK_RESULT, r);
      return;
    }
  
  }

  /* traceback word trellis from the best word */
  total_lscore = trace_backptr(wordseq, &wordlen, best, r->lm->winfo);
#ifdef SPSEGMENT_NAIST
  if (r->config->successive.enabled) {
    /* on segmentation mode, recognition result that only consists of
       short-pause words will be treated as recognition rejection */
    ok_p = FALSE;
    for(i=0;i<wordlen;i++) {
      if (! is_sil(wordseq[i], r)) ok_p = TRUE;
    }
    if (ok_p == FALSE) {
      r->result.status = J_RESULT_STATUS_ONLY_SILENCE;
      return;
    }
  }
#endif

  /* just flush last progress output */
  /*
  if (recog->jconf->output.progout_flag) {
    recog->result.status = 1;
    recog->result.num_frame = last_time;
    recog->result.pass1.word = wordseq;
    recog->result.pass1.word_num = wordlen;
    recog->result.pass1.score = best->backscore;
    recog->result.pass1.score_lm = total_lscore;
    recog->result.pass1.score_am = best->backscore - total_lscore;
    //callback_exec(CALLBACK_RESULT_PASS1_INTERIM, recog);
    }*/

  /* output 1st pass result */    
  if (verbose_flag || ! r->config->output.progout_flag) {
    r->result.status = J_RESULT_STATUS_SUCCESS;
    r->result.num_frame = framelen;
    for(i=0;i<wordlen;i++) r->result.pass1.word[i] = wordseq[i];
    r->result.pass1.word_num = wordlen;
    r->result.pass1.score = best->backscore;
    r->result.pass1.score_lm = total_lscore;
    r->result.pass1.score_am = best->backscore - total_lscore;
    //callback_exec(CALLBACK_RESULT_PASS1, r);
  }

  /* store the result to global val (notice: in reverse order) */
  for(i=0;i<wordlen;i++) r->pass1_wseq[i] = wordseq[i];
  r->pass1_wnum = wordlen;
  r->pass1_score = best->backscore;

#ifdef WORD_GRAPH
  /* 帽胳トレリスから·ラティスを栏喇する */
  /* generate word graph from the word trellis */
  r->peseqlen = backtrellis->framelen;
  r->result.wg1 = NULL;
  generate_lattice(last_time, r);
  link_lattice_by_time(r->result.wg1);
  if (r->lmtype == LM_PROB) re_compute_lattice_lm(r->result.wg1, r->wchmm);
  r->result.wg1_num = wordgraph_sort_and_annotate_id(&(r->result.wg1), r);
  /* compute graph CM by forward-backward processing */
  graph_forward_backward(r->result.wg1, r);
  //callback_exec(CALLBACK_RESULT_PASS1_GRAPH, r);
  //wordgraph_clean(&(r->result.wg1));
#endif

}

/** 
 * <JA>
 * トレリス帽胳をスコアでソ〖トするqsort簇眶. 
 * 
 * @param x1 [in] 妥燎1へのポインタ
 * @param x2 [in] 妥燎2へのポインタ
 * 
 * @return qsort の猛
 * </JA>
 * <EN>
 * qsort function to sort trellis words by their score.
 * 
 * @param x1 [in] pointer to element #1
 * @param x2 [in] pointer to element #2
 * 
 * @return value required for qsort.
 * </EN>
 */
static int
compare_backscore(TRELLIS_ATOM **x1, TRELLIS_ATOM **x2)
{
  return((*x2)->backscore - (*x1)->backscore);
}

/** 
 * <JA>
 * find_1pass_result() の帽胳千急モ〖ド惹. 帽胳千急モ〖ドでは妈1パスで
 * 千急を姜位するので·评られた铬输は奶撅の妈２パスと票じ眷疥に呈羌する. 
 * 
 * @param framelen [in] 妈１パスで借妄が毗茫したフレ〖ム眶
 * @param r [i/o] 千急借妄インスタンス
 * 
 * </JA>
 * <EN>
 * Isolated word recognition version of find_1pass_result().
 * Since Julius executes only the 1st pass on Isolated word recognition
 * mode, the result candidate will be stored as the final result.
 * 
 * @param framelen [in] frame length that has been processed
 * @param r [i/o] recognition process instance
 * 
 * </EN>
 */
static void
find_1pass_result_word(int framelen, RecogProcess *r)
{
  BACKTRELLIS *bt;
  TRELLIS_ATOM *best, *tmp;
  int last_time;
  Sentence *s;
#ifdef CONFIDENCE_MEASURE
  LOGPROB sum;
#endif
  LOGPROB maxscore;
  int i;
  TRELLIS_ATOM **idx;
  int num;

  if (r->lmvar != LM_DFA_WORD) return;

  bt = r->backtrellis;
  for (last_time = framelen - 1; last_time >= 0; last_time--) {
    maxscore = LOG_ZERO;
    for (i=0;i<bt->num[last_time];i++) {
      tmp = bt->rw[last_time][i];
#ifdef WORD_GRAPH
      /* treat only words on a graph path */
      if (!tmp->within_context) continue;
#endif
      if (maxscore < tmp->backscore) {
	maxscore = tmp->backscore;
	best = tmp;
      }
    }
    if (maxscore != LOG_ZERO) break;
  }

  if (last_time < 0) {		/* not found */
    jlog("WARNING: %02d %s: no word survived on the last frame, search failed\n", r->config->id, r->config->name);
    r->result.status = J_RESULT_STATUS_FAIL;
    //callback_exec(CALLBACK_RESULT, r);
    return;
  }

#ifdef CONFIDENCE_MEASURE
  sum = 0.0;
  for (i=0;i<bt->num[last_time];i++) {
    tmp = bt->rw[last_time][i];
#ifdef WORD_GRAPH
    /* treat only words on a graph path */
    if (!tmp->within_context) continue;
#endif
    sum += pow(10, r->config->annotate.cm_alpha * (tmp->backscore - maxscore));
  }
#endif

  /* set recognition result status to normal */
  r->result.status = J_RESULT_STATUS_SUCCESS;

  if (r->config->output.output_hypo_maxnum > 1) {
    /* more than one candidate is requested */

    /* get actual number of candidates to output */
    num = r->config->output.output_hypo_maxnum;
    if (num > bt->num[last_time]) {
      num = bt->num[last_time];
    }

    /* prepare result storage */
    result_sentence_malloc(r, num);
    r->result.sentnum = num;

    /* sort by score */
    idx = (TRELLIS_ATOM **)mymalloc(sizeof(TRELLIS_ATOM *)*bt->num[last_time]);
    for (i=0;i<bt->num[last_time];i++) {
      idx[i] = bt->rw[last_time][i];
    }
    qsort(idx, bt->num[last_time], sizeof(TRELLIS_ATOM *),
	  (int (*)(const void *,const void *))compare_backscore);
    
    /* store to result storage */
    for(i=0;i<r->result.sentnum;i++) {
      s = &(r->result.sent[i]);
      tmp = idx[i];
      s->word_num = 1;
      s->word[0] = tmp->wid;
#ifdef CONFIDENCE_MEASURE
      s->confidence[0] = pow(10, r->config->annotate.cm_alpha * (tmp->backscore - maxscore)) / sum;
#endif
      s->score = tmp->backscore;
      s->score_lm = 0.0;
      s->score_am = tmp->backscore;
      if (multigram_get_all_num(r->lm) > 0) {
	s->gram_id = multigram_get_gram_from_wid(s->word[0], r->lm);
      } else {
	s->gram_id = 0;
      }
    }
    /* free work area for sort */
    free(idx);

  } else {			/* only max is needed */

    /* prepare result storage */
    result_sentence_malloc(r, 1);
    r->result.sentnum = 1;
    s = &(r->result.sent[0]);
    s->word_num = 1;
    s->word[0] = best->wid;
#ifdef CONFIDENCE_MEASURE
    s->confidence[0] = 1.0 / sum;
#endif
    s->score = best->backscore;
    s->score_lm = 0.0;
    s->score_am = best->backscore;
    if (multigram_get_all_num(r->lm) > 0) {
      s->gram_id = multigram_get_gram_from_wid(s->word[0], r->lm);
    } else {
      s->gram_id = 0;
    }
  }

  /* copy as 1st pass result */
  memcpy(&(r->result.pass1), &(r->result.sent[0]), sizeof(Sentence));
  r->result.pass1.align = NULL;

  //callback_exec(CALLBACK_RESULT, r);
  //free(r->result.sent);
}


#ifdef DETERMINE

/** 
 * <JA>
 * 妈１パスの庞面デ〖タから玲袋澄年材墙かどうか冉年する∈悸赋∷. tremax が
 * NULL のときは介袋步する. 澄年箕は r->have_determine を TRUE にする. 
 *
 * @param r [i/o] 不兰千急借妄インスタンス
 * @param t [in] フレ〖ム
 * @param tremax [in] 附哼のフレ〖ム惧で呵锑のトレリス帽胳
 * @param thres [in] 澄年脱のスコア镧猛
 * @param countthres [in] 澄年脱の积鲁フレ〖ム眶の镧猛
 *
 * @return 澄年箕は tremax を手す. 踏澄年箕は NULL を手す. 
 * </JA>
 * <EN>
 * Try to Determine a word hypothesis before end of input on isolated
 * word recognition mode (EXPERIMENT).  Initialize if tremax is NULL.
 * Set r->have_determine to TRUE if determined.
 * 
 * @param r [i/o] recognition process instance
 * @param t [in] current frame
 * @param tremax [in] maximum scored trellis word on the current frame
 * @param thres [in] score threshold for determinization
 * @param countthres [in] frame duration threshold for determinization
 *
 * @return the tremax if determined, or NULL if not determined yet., 
 * </EN>
 */
static TRELLIS_ATOM *
determine_word(RecogProcess *r, int t, TRELLIS_ATOM *tremax, LOGPROB thres, int countthres)
{
  TRELLIS_ATOM *ret;
  WORD_ID w;

  //LOGPROB sum;
  //LOGPROB cm;

  int j;
  FSBeam *d;
  TOKEN2 *tk;
    
  if (tremax == NULL) {
    /* initialize */
    r->determine_count = 0;
    r->determine_maxnodescore = LOG_ZERO;
    r->determined = FALSE;
    r->determine_last_wid = WORD_INVALID;
    r->have_determine = FALSE;
    return NULL;
  }

  ret = NULL;

  /* get confidence score of the maximum word hypothesis */
/* 
 *   sum = 0.0;
 *   tre = recog->backtrellis->list;
 *   while (tre != NULL && tre->endtime == t) {
 *     sum += pow(10, recog->jconf->annotate.cm_alpha * (tre->backscore - tremax->backscore));
 *     tre = tre->next;
 *   }
 *   cm = 1.0 / sum;
 */

  /* determinization decision */
  w = tremax->wid;

  r->have_determine = FALSE;

  /* determine by score threshold from maximum node score to maximum word end node score */
  if (r->determine_last_wid == w && r->determine_maxnodescore - tremax->backscore <= thres) {
    r->determine_count++;
    if (r->determine_count > countthres) {
      if (r->determined == FALSE) {
	ret = tremax;
	r->determined = TRUE;
	r->have_determine = TRUE;
      }
    }
  } else {
    r->determine_count = 0;
  }

  //printf("determine: %d: %s: cm=%f, relscore=%f, count=%d, phase=%d\n", t, recog->model->winfo->woutput[w], cm, determine_maxnodescore - tremax->backscore, count, phase);
  r->determine_last_wid = w;

  /* update maximum node score here for next call, since
     the word path determination is always one frame later */
  d = &(r->pass1);
  r->determine_maxnodescore = LOG_ZERO;
  for (j = d->n_start; j <= d->n_end; j++) {
    tk = &(d->tlist[d->tn][d->tindex[d->tn][j]]);
    if (r->determine_maxnodescore < tk->score) r->determine_maxnodescore = tk->score;
  }

  return(ret);
}

/** 
 * <JA>
 * 帽胳千急箕に·妈１パスの借妄面に玲袋澄年する∈悸赋∷. 澄年できた眷圭·