libjulius/src/word_align.c File Reference

Forced alignment by word / phoneme / state unit. More...

#include <julius/julius.h>

Go to the source code of this file.

Functions
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)
	Make the whole sentence HMM from given word sequence by connecting each phoneme HMM.
static void	do_align (WORD_ID *words, short wnum, HTK_Param *param, int per_what, Sentence *s, RecogProcess *r)
	Build sentence HMM, call viterbi_segment() and output result.
void	word_align (WORD_ID *words, short wnum, HTK_Param *param, Sentence *s, RecogProcess *r)
	Do forced alignment per word for the given word sequence.
void	word_rev_align (WORD_ID *revwords, short wnum, HTK_Param *param, Sentence *s, RecogProcess *r)
	Do forced alignment per word for the given word sequence (reversed order).
void	phoneme_align (WORD_ID *words, short num, HTK_Param *param, Sentence *s, RecogProcess *r)
	Do forced alignment per phoneme for the given word sequence.
void	phoneme_rev_align (WORD_ID *revwords, short num, HTK_Param *param, Sentence *s, RecogProcess *r)
	Do forced alignment per phoneme for the given word sequence (reversed order).
void	state_align (WORD_ID *words, short num, HTK_Param *param, Sentence *s, RecogProcess *r)
	Do forced alignment per HMM state for the given word sequence.
void	state_rev_align (WORD_ID *revwords, short num, HTK_Param *param, Sentence *s, RecogProcess *r)
	Do forced alignment per state for the given word sequence (reversed order).

---

Detailed Description

Forced alignment by word / phoneme / state unit.

This file defines functions for performing forced alignment of recognized words. The forced alignment is implimented in Julius/Julian to get the best matching segmentation of recognized word sequence upon input speech. Word-level, phoneme-level and HMM state-level alignment can be obtained.

Julius/Julian performs the forced alignment as a post-processing of recognition process. Recomputation of Viterbi path on the recognized word sequence toward input speech will be done after the recognition to get better alignment.

Author:

Akinobu Lee

Date:

Sat Sep 24 16:09:46 2005

Revision

1.1.1.1

Definition in file word_align.c.

---

Function Documentation

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
	)			[static]

Make the whole sentence HMM from given word sequence by connecting each phoneme HMM.

Parameters:

	wseq	[in] word sequence to align
	num	[in] number of wseq
	has_sp_ret	[out] unit information of whether it can be followed by a short-pause
	num_ret	[out] number of HMM contained in the generated sentence HMM
	end_ret	[out] sequence of state location as alignment unit
	per_what	[in] specify the alignment unit (word / phoneme / state)
	r	[in] recognition process instance

Returns:

newly malloced HMM sequences.

Definition at line 78 of file word_align.c.

Referenced by do_align().

static void do_align	(	WORD_ID *	words,
		short	wnum,
		HTK_Param *	param,
		int	per_what,
		Sentence *	s,
		RecogProcess *	r
	)			[static]

Build sentence HMM, call viterbi_segment() and output result.

Parameters:

	words	[in] word sequence of the sentence
	wnum	[in] number of words in words
	param	[in] input parameter vector
	per_what	[in] specify the alignment unit (word / phoneme / state)
	s	[out] Sentence data area to store the alignment result
	r	[i/o] recognition process instance

Definition at line 186 of file word_align.c.

Referenced by phoneme_align(), phoneme_rev_align(), state_align(), state_rev_align(), word_align(), and word_rev_align().

void word_align	(	WORD_ID *	words,
		short	wnum,
		HTK_Param *	param,
		Sentence *	s,
		RecogProcess *	r
	)

Do forced alignment per word for the given word sequence.

Parameters:

	words	[in] word sequence
	wnum	[in] length of words
	param	[in] input parameter vectors
	s	[out] Sentence data area to store the alignment result
	r	[i/o] recognition process instance

Definition at line 370 of file word_align.c.

void word_rev_align	(	WORD_ID *	revwords,
		short	wnum,
		HTK_Param *	param,
		Sentence *	s,
		RecogProcess *	r
	)

Do forced alignment per word for the given word sequence (reversed order).

Parameters:

	revwords	[in] word sequence in reversed direction
	wnum	[in] length of revwords
	param	[in] input parameter vectors
	s	[out] Sentence data area to store the alignment result
	r	[i/o] recognition process instance

Definition at line 398 of file word_align.c.

void phoneme_align	(	WORD_ID *	words,
		short	num,
		HTK_Param *	param,
		Sentence *	s,
		RecogProcess *	r
	)

Do forced alignment per phoneme for the given word sequence.

Parameters:

	words	[in] word sequence
	num	[in] length of words
	param	[in] input parameter vectors
	s	[out] Sentence data area to store the alignment result
	r	[i/o] recognition process instance

Definition at line 431 of file word_align.c.

void phoneme_rev_align	(	WORD_ID *	revwords,
		short	num,
		HTK_Param *	param,
		Sentence *	s,
		RecogProcess *	r
	)

Do forced alignment per phoneme for the given word sequence (reversed order).

Parameters:

	revwords	[in] word sequence in reversed direction
	num	[in] length of revwords
	param	[in] input parameter vectors
	s	[out] Sentence data area to store the alignment result
	r	[i/o] recognition process instance

Definition at line 459 of file word_align.c.

void state_align	(	WORD_ID *	words,
		short	num,
		HTK_Param *	param,
		Sentence *	s,
		RecogProcess *	r
	)

Do forced alignment per HMM state for the given word sequence.

Parameters:

	words	[in] word sequence
	num	[in] length of words
	param	[in] input parameter vectors
	s	[out] Sentence data area to store the alignment result
	r	[i/o] recognition process instance

Definition at line 492 of file word_align.c.

void state_rev_align	(	WORD_ID *	revwords,
		short	num,
		HTK_Param *	param,
		Sentence *	s,
		RecogProcess *	r
	)

Do forced alignment per state for the given word sequence (reversed order).

Parameters:

	revwords	[in] word sequence in reversed direction
	num	[in] length of revwords
	param	[in] input parameter vectors
	s	[out] Sentence data area to store the alignment result
	r	[i/o] recognition process instance

Definition at line 520 of file word_align.c.

---