libjulius/src/word_align.c

Go to the documentation of this file.

/*
 * Copyright (c) 1991-2007 Kawahara Lab., Kyoto University
 * Copyright (c) 2000-2005 Shikano Lab., Nara Institute of Science and Technology
 * Copyright (c) 2005-2007 Julius project team, Nagoya Institute of Technology
 * All rights reserved
 */

#include <julius/julius.h>

static HMM_Logical **
make_phseq(WORD_ID *wseq, short num, boolean **has_sp_ret, int *num_ret, int **end_ret, int per_what, 
           RecogProcess *r)
{
  HMM_Logical **ph;             /* phoneme sequence */
  boolean *has_sp;
  int k;
  int phnum;                    /* num of above */
  WORD_ID tmpw, w;
  int i, j, pn, st, endn;
  HMM_Logical *tmpp, *ret;
  WORD_INFO *winfo;
  HTK_HMM_INFO *hmminfo;
  boolean enable_iwsp;          /* for multipath */

  winfo = r->lm->winfo;
  hmminfo = r->am->hmminfo;
  if (hmminfo->multipath) enable_iwsp = r->lm->config->enable_iwsp;

  /* make ph[] from wseq[] */
  /* 1. calc total phone num and malloc */
  phnum = 0;
  for (w=0;w<num;w++) phnum += winfo->wlen[wseq[w]];
  ph = (HMM_Logical **)mymalloc(sizeof(HMM_Logical *) * phnum);
  
  if (hmminfo->multipath) {
    has_sp = (boolean *)mymalloc(sizeof(boolean) * phnum);
  }
  /* 2. make phoneme sequence */
  st = 0;
  if (hmminfo->multipath) st++;
  pn = 0;
  endn = 0;
  for (w=0;w<num;w++) {
    tmpw = wseq[w];
    for (i=0;i<winfo->wlen[tmpw];i++) {
      tmpp = winfo->wseq[tmpw][i];
      /* handle cross-word context dependency */
      if (r->ccd_flag) {
        if (w > 0 && i == 0) {  /* word head */
          
          if ((ret = get_left_context_HMM(tmpp, ph[pn-1]->name, hmminfo)) != NULL) {
            tmpp = ret;
          }
          /* if triphone not found, fallback to bi/mono-phone  */
          /* use pseudo phone when no bi-phone found in alignment... */
        }
        if (w < num-1 && i == winfo->wlen[tmpw] - 1) { /* word tail */
          if ((ret = get_right_context_HMM(tmpp, winfo->wseq[wseq[w+1]][0]->name, hmminfo)) != NULL) {
            tmpp = ret;
          }
        }
      }
      ph[pn] = tmpp;
      if (hmminfo->multipath) {
        if (enable_iwsp && i == winfo->wlen[tmpw] - 1) {
          has_sp[pn] = TRUE;
        } else {
          has_sp[pn] = FALSE;
        }
      }
      if (per_what == PER_STATE) {
        for (j=0;j<hmm_logical_state_num(tmpp)-2;j++) {
          (*end_ret)[endn++] = st + j;
        }
        if (hmminfo->multipath && enable_iwsp && has_sp[pn]) {
          for (k=0;k<hmm_logical_state_num(hmminfo->sp)-2;k++) {
            (*end_ret)[endn++] = st + j + k;
          }
        }
      }
      st += hmm_logical_state_num(tmpp) - 2;
      if (hmminfo->multipath && enable_iwsp && has_sp[pn]) {
        st += hmm_logical_state_num(hmminfo->sp) - 2;
      }
      if (per_what == PER_PHONEME) (*end_ret)[endn++] = st - 1;
      pn++;
    }
    if (per_what == PER_WORD) (*end_ret)[endn++] = st - 1;
  }
  *num_ret = phnum;
  if (hmminfo->multipath) *has_sp_ret = has_sp;
  return ph;
}


static void
do_align(WORD_ID *words, short wnum, HTK_Param *param, int per_what, Sentence *s, RecogProcess *r)
{
  HMM_Logical **phones;         /* phoneme sequence */
  boolean *has_sp;              /* whether phone can follow short pause */
  int k;
  int phonenum;                 /* num of above */
  HMM *shmm;                    /* sentence HMM */
  int *end_state;               /* state number of word ends */
  int *end_frame;               /* segmented last frame of words */
  LOGPROB *end_score;           /* normalized score of each words */
  LOGPROB allscore;             /* total score of this word sequence */
  WORD_ID w;
  int i, rlen;
  int end_num = 0;
  int *id_seq, *phloc = NULL, *stloc = NULL;
  int j,n,p;
  WORD_INFO *winfo;
  HTK_HMM_INFO *hmminfo;
  boolean enable_iwsp;          /* for multipath */

  winfo = r->lm->winfo;
  hmminfo = r->am->hmminfo;
  if (hmminfo->multipath) enable_iwsp = r->lm->config->enable_iwsp;

  /* initialize result storage buffer */
  switch(per_what) {
  case PER_WORD:
    jlog("ALIGN: === word alignment begin ===\n");
    end_num = wnum;
    phloc = (int *)mymalloc(sizeof(int)*wnum);
    i = 0;
    for(w=0;w<wnum;w++) {
      phloc[w] = i;
      i += winfo->wlen[words[w]];
    }
    break;
  case PER_PHONEME:
    jlog("ALIGN: === phoneme alignment begin ===\n");
    end_num = 0;
    for(w=0;w<wnum;w++) end_num += winfo->wlen[words[w]];
    break;
  case PER_STATE:
    jlog("ALIGN: === state alignment begin ===\n");
    end_num = 0;
    for(w=0;w<wnum;w++) {
      for (i=0;i<winfo->wlen[words[w]]; i++) {
        end_num += hmm_logical_state_num(winfo->wseq[words[w]][i]) - 2;
      }
      if (hmminfo->multipath && enable_iwsp) {
        end_num += hmm_logical_state_num(hmminfo->sp) - 2;
      }
    }
    phloc = (int *)mymalloc(sizeof(int)*end_num);
    stloc = (int *)mymalloc(sizeof(int)*end_num);
    {
      n = 0;
      p = 0;
      for(w=0;w<wnum;w++) {
        for(i=0;i<winfo->wlen[words[w]]; i++) {
          for(j=0; j<hmm_logical_state_num(winfo->wseq[words[w]][i]) - 2; j++) {
            phloc[n] = p;
            stloc[n] = j + 1;
            n++;
          }
          if (hmminfo->multipath && enable_iwsp && i == winfo->wlen[words[w]] - 1) {
            for(k=0;k<hmm_logical_state_num(hmminfo->sp)-2;k++) {
              phloc[n] = p;
              stloc[n] = j + 1 + k + end_num;
              n++;
            }
          }
          p++;
        }
      }
    }
    
    break;
  }
  end_state = (int *)mymalloc(sizeof(int) * end_num);

  /* make phoneme sequence word sequence */
  phones = make_phseq(words, wnum, hmminfo->multipath ? &has_sp : NULL, &phonenum, &end_state, per_what, r);
  /* build the sentence HMMs */
  shmm = new_make_word_hmm(hmminfo, phones, phonenum, hmminfo->multipath ? has_sp : NULL);
  if (shmm == NULL) {
    j_internal_error("Error: failed to make word hmm for alignment\n");
  }

  /* call viterbi segmentation function */
  allscore = viterbi_segment(shmm, param, r->wchmm->hmmwrk, hmminfo->multipath, end_state, end_num, &id_seq, &end_frame, &end_score, &rlen);

  /* store result to s */
  s->align.num = rlen;
  s->align.unittype = per_what;
  s->align.begin_frame = (int *)mymalloc(sizeof(int) * rlen);
  s->align.end_frame   = (int *)mymalloc(sizeof(int) * rlen);
  s->align.avgscore    = (LOGPROB *)mymalloc(sizeof(LOGPROB) * rlen);
  for(i=0;i<rlen;i++) {
    s->align.begin_frame[i] = (i == 0) ? 0 : end_frame[i-1] + 1;
    s->align.end_frame[i]   = end_frame[i];
    s->align.avgscore[i]    = end_score[i];
  }
  switch(per_what) {
  case PER_WORD:
    s->align.w = (WORD_ID *)mymalloc(sizeof(WORD_ID) * rlen);
    for(i=0;i<rlen;i++) {
      s->align.w[i] = words[id_seq[i]];
    }
    break;
  case PER_PHONEME:
    s->align.ph = (HMM_Logical **)mymalloc(sizeof(HMM_Logical *) * rlen);
    for(i=0;i<rlen;i++) {
      s->align.ph[i] = phones[id_seq[i]];
    }
    break;
  case PER_STATE:
    s->align.ph = (HMM_Logical **)mymalloc(sizeof(HMM_Logical *) * rlen);
    s->align.loc = (short *)mymalloc(sizeof(short) * rlen);
    if (hmminfo->multipath) s->align.is_iwsp = (boolean *)mymalloc(sizeof(boolean) * rlen);
    for(i=0;i<rlen;i++) {
      s->align.ph[i]  = phones[phloc[id_seq[i]]];
      if (hmminfo->multipath) {
        if (enable_iwsp && stloc[id_seq[i]] > end_num) {
          s->align.loc[i] = stloc[id_seq[i]] - end_num;
          s->align.is_iwsp[i] = TRUE;
        } else {
          s->align.loc[i] = stloc[id_seq[i]];
          s->align.is_iwsp[i] = FALSE;
        }
      } else {
        s->align.loc[i] = stloc[id_seq[i]];
      }
    }
    break;
  }

  s->align.allscore = allscore;

  s->align.filled = TRUE;

  free_hmm(shmm);
  free(id_seq);
  free(phones);
  if (hmminfo->multipath) free(has_sp);
  free(end_score);
  free(end_frame);
  free(end_state);

  switch(per_what) {
  case PER_WORD:
    free(phloc);
    break;
  case PER_PHONEME:
    break;
  case PER_STATE:
    free(phloc);
    free(stloc);
  }
  
}

void
word_align(WORD_ID *words, short wnum, HTK_Param *param, Sentence *s, RecogProcess *r)
{
  do_align(words, wnum, param, PER_WORD, s, r);
}

void
word_rev_align(WORD_ID *revwords, short wnum, HTK_Param *param, Sentence *s, RecogProcess *r)
{
  WORD_ID *words;               /* word sequence (true order) */
  int w;
  words = (WORD_ID *)mymalloc(sizeof(WORD_ID) * wnum);
  for (w=0;w<wnum;w++) words[w] = revwords[wnum-w-1];
  do_align(words, wnum, param, PER_WORD, s, r);
  free(words);
}

void
phoneme_align(WORD_ID *words, short num, HTK_Param *param, Sentence *s, RecogProcess *r)
{
  do_align(words, num, param, PER_PHONEME, s, r);
}

void
phoneme_rev_align(WORD_ID *revwords, short num, HTK_Param *param, Sentence *s, RecogProcess *r)
{
  WORD_ID *words;               /* word sequence (true order) */
  int p;
  words = (WORD_ID *)mymalloc(sizeof(WORD_ID) * num);
  for (p=0;p<num;p++) words[p] = revwords[num-p-1];
  do_align(words, num, param, PER_PHONEME, s, r);
  free(words);
}

void
state_align(WORD_ID *words, short num, HTK_Param *param, Sentence *s, RecogProcess *r)
{
  do_align(words, num, param, PER_STATE, s, r);
}

void
state_rev_align(WORD_ID *revwords, short num, HTK_Param *param, Sentence *s, RecogProcess *r)
{
  WORD_ID *words;               /* word sequence (true order) */
  int p;
  words = (WORD_ID *)mymalloc(sizeof(WORD_ID) * num);
  for (p=0;p<num;p++) words[p] = revwords[num-p-1];
  do_align(words, num, param, PER_STATE, s, r);
  free(words);
}

/* end of file */

---