word_align.c

julius version 4.12.about sound recognition.

    align->w = (WORD_ID *)mymalloc(sizeof(WORD_ID) * rlen);
    for(i=0;i<rlen;i++) {
      align->w[i] = words[id_seq[i]];
    }
    break;
  case PER_PHONEME:
    align->ph = (HMM_Logical **)mymalloc(sizeof(HMM_Logical *) * rlen);
    for(i=0;i<rlen;i++) {
      align->ph[i] = phones[id_seq[i]];
    }
    break;
  case PER_STATE:
    align->ph = (HMM_Logical **)mymalloc(sizeof(HMM_Logical *) * rlen);
    align->loc = (short *)mymalloc(sizeof(short) * rlen);
    if (hmminfo->multipath) align->is_iwsp = (boolean *)mymalloc(sizeof(boolean) * rlen);
    for(i=0;i<rlen;i++) {
      align->ph[i]  = phones[phloc[id_seq[i]]];
      if (hmminfo->multipath) {
	if (enable_iwsp && stloc[id_seq[i]] > end_num) {
	  align->loc[i] = stloc[id_seq[i]] - end_num;
	  align->is_iwsp[i] = TRUE;
	} else {
	  align->loc[i] = stloc[id_seq[i]];
	  align->is_iwsp[i] = FALSE;
	}
      } else {
	align->loc[i] = stloc[id_seq[i]];
      }
    }
    break;
  }

  align->allscore = allscore;

  free_hmm(shmm);
  free(id_seq);
  free(phones);
  if (has_sp) 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);
  }
  
}

/** 
 * <JA>
 * 帽胳ごとの forced alignment を乖う. 
 * 
 * @param words [in] 帽胳误
 * @param wnum [in] @a words の帽胳眶
 * @param param [in] 掐蜗泼魔ベクトル误
 * @param align [out] アラインメント冯蔡を呈羌するSentence菇陇挛
 * @param r [i/o] 千急借妄インスタンス
 * </JA>
 * <EN>
 * Do forced alignment per word for the given word sequence.
 * 
 * @param words [in] word sequence
 * @param wnum [in] length of @a words
 * @param param [in] input parameter vectors
 * @param align [out] Sentence data area to store the alignment result
 * @param r [i/o] recognition process instance
 * </EN>
 * @callgraph
 * @callergraph
 */
void
word_align(WORD_ID *words, short wnum, HTK_Param *param, SentenceAlign *align, RecogProcess *r)
{
  do_align(words, wnum, param, PER_WORD, align, r);
}

/** 
 * <JA>
 * 帽胳ごとの forced alignment を乖う∈帽胳が嫡界で涂えられる眷圭∷
 * 
 * @param revwords [in] 帽胳误∈嫡界∷
 * @param wnum [in] @a revwords の帽胳眶
 * @param param [in] 掐蜗泼魔ベクトル误
 * @param align [out] アラインメント冯蔡を呈羌するSentence菇陇挛
 * @param r [i/o] 千急借妄インスタンス
 * </JA>
 * <EN>
 * Do forced alignment per word for the given word sequence (reversed order).
 * 
 * @param revwords [in] word sequence in reversed direction
 * @param wnum [in] length of @a revwords
 * @param param [in] input parameter vectors
 * @param align [out] Sentence data area to store the alignment result
 * @param r [i/o] recognition process instance
 * </EN>
 * @callgraph
 * @callergraph
 */
void
word_rev_align(WORD_ID *revwords, short wnum, HTK_Param *param, SentenceAlign *align, 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, align, r);
  free(words);
}

/** 
 * <JA>
 * 不燎ごとの forced alignment を乖う. 
 * 
 * @param words [in] 帽胳误
 * @param num [in] @a words の帽胳眶
 * @param param [in] 掐蜗泼魔ベクトル误
 * @param align [out] アラインメント冯蔡を呈羌するSentence菇陇挛
 * @param r [i/o] 千急借妄インスタンス
 * </JA>
 * <EN>
 * Do forced alignment per phoneme for the given word sequence.
 * 
 * @param words [in] word sequence
 * @param num [in] length of @a words
 * @param param [in] input parameter vectors
 * @param align [out] Sentence data area to store the alignment result
 * @param r [i/o] recognition process instance
 * </EN>
 * @callgraph
 * @callergraph
 */
void
phoneme_align(WORD_ID *words, short num, HTK_Param *param, SentenceAlign *align, RecogProcess *r)
{
  do_align(words, num, param, PER_PHONEME, align, r);
}

/** 
 * <JA>
 * 不燎ごとの forced alignment を乖う∈帽胳が嫡界で涂えられる眷圭∷
 * 
 * @param revwords [in] 帽胳误∈嫡界∷
 * @param num [in] @a revwords の帽胳眶
 * @param param [in] 掐蜗泼魔ベクトル误
 * @param align [out] アラインメント冯蔡を呈羌するSentence菇陇挛
 * @param r [i/o] 千急借妄インスタンス
 * </JA>
 * <EN>
 * Do forced alignment per phoneme for the given word sequence (reversed order).
 * 
 * @param revwords [in] word sequence in reversed direction
 * @param num [in] length of @a revwords
 * @param param [in] input parameter vectors
 * @param align [out] Sentence data area to store the alignment result
 * @param r [i/o] recognition process instance
 * </EN>
 * @callgraph
 * @callergraph
 */
void
phoneme_rev_align(WORD_ID *revwords, short num, HTK_Param *param, SentenceAlign *align, 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, align, r);
  free(words);
}

/** 
 * <JA>
 * HMM觉轮ごとの forced alignment を乖う. 
 * 
 * @param words [in] 帽胳误
 * @param num [in] @a words の帽胳眶
 * @param param [in] 掐蜗泼魔ベクトル误
 * @param align [out] アラインメント冯蔡を呈羌するSentence菇陇挛
 * @param r [i/o] 千急借妄インスタンス
 * </JA>
 * <EN>
 * Do forced alignment per HMM state for the given word sequence.
 * 
 * @param words [in] word sequence
 * @param num [in] length of @a words
 * @param param [in] input parameter vectors
 * @param align [out] Sentence data area to store the alignment result
 * @param r [i/o] recognition process instance
 * </EN>
 * @callgraph
 * @callergraph
 */
void
state_align(WORD_ID *words, short num, HTK_Param *param, SentenceAlign *align, RecogProcess *r)
{
  do_align(words, num, param, PER_STATE, align, r);
}

/** 
 * <JA>
 * HMM觉轮ごとの forced alignment を乖う∈帽胳が嫡界で涂えられる眷圭∷
 * 
 * @param revwords [in] 帽胳误∈嫡界∷
 * @param num [in] @a revwords の帽胳眶
 * @param param [in] 掐蜗泼魔ベクトル误
 * @param align [out] アラインメント冯蔡を呈羌するSentence菇陇挛
 * @param r [i/o] 千急借妄インスタンス
 * </JA>
 * <EN>
 * Do forced alignment per state for the given word sequence (reversed order).
 * 
 * @param revwords [in] word sequence in reversed direction
 * @param num [in] length of @a revwords
 * @param param [in] input parameter vectors
 * @param align [out] Sentence data area to store the alignment result
 * @param r [i/o] recognition process instance
 * </EN>
 * @callgraph
 * @callergraph
 */
void
state_rev_align(WORD_ID *revwords, short num, HTK_Param *param, SentenceAlign *align, 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, align, r);
  free(words);
}

/** 
 * <JA>
 * 千急冯蔡に滦して涩妥なアラインメントを链て悸乖するˉ
 * 
 * @param r [i/o] 千急借妄インスタンス
 * @param param [in] 掐蜗泼魔ベクトル误
 * </JA>
 * <EN>
 * Do required forced alignment for the recognition results
 * 
 * @param r [i/o] recognition process instance
 * @param param [in] input parameter vectors
 * </EN>
 * @callgraph
 * @callergraph
 */
void
do_alignment_all(RecogProcess *r, HTK_Param *param)
{
  int n;
  Sentence *s;
  SentenceAlign *now, *prev;

  for(n = 0; n < r->result.sentnum; n++) {
    s = &(r->result.sent[n]);
    /* do forced alignment if needed */
    if (r->config->annotate.align_result_word_flag) {
      now = result_align_new();
      word_align(s->word, s->word_num, param, now, r);
      if (s->align == NULL) s->align = now;
      else prev->next = now;
      prev = now;
    }
    if (r->config->annotate.align_result_phoneme_flag) {
      now = result_align_new();
      phoneme_align(s->word, s->word_num, param, now, r);
      if (s->align == NULL) s->align = now;
      else prev->next = now;
      prev = now;
    }
    if (r->config->annotate.align_result_state_flag) {
      now = result_align_new();
      state_align(s->word, s->word_num, param, now, r);
      if (s->align == NULL) s->align = now;
      else prev->next = now;
      prev = now;
    }
  }
} 

/* end of file */