outprob.c File Reference

Computation of acoustic likelihood in HMM state, with state-level cache. More...

#include <sent/stddefs.h>
#include <sent/speech.h>
#include <sent/htk_hmm.h>
#include <sent/htk_param.h>
#include <sent/hmm.h>
#include <sent/gprune.h>
#include "globalvars.h"

Include dependency graph for outprob.c:

Go to the source code of this file.

Defines
#define	LOG_UNDEF   (LOG_ZERO - 1)
	Value to be used as the initial cache value.
Functions
boolean	outprob_cache_init ()
boolean	outprob_cache_prepare ()
static void	outprob_cache_extend (int reqframe)
LOGPROB	outprob_state (int t, HTK_HMM_State *stateinfo, HTK_Param *param)
	Compute output probability of a state.
void	outprob_cd_nbest_init (int num)
static LOGPROB	outprob_cd_nbest (int t, CD_State_Set *lset, HTK_Param *param)
static LOGPROB	outprob_cd_max (int t, CD_State_Set *lset, HTK_Param *param)
static LOGPROB	outprob_cd_avg (int t, CD_State_Set *lset, HTK_Param *param)
LOGPROB	outprob_cd (int t, CD_State_Set *lset, HTK_Param *param)
LOGPROB	outprob (int t, HMM_STATE *hmmstate, HTK_Param *param)
Variables
static int	statenum
	Local work area that holds total number of HMM states in the HMM definition data.
static LOGPROB **	outprob_cache = NULL
	State-level cache [t][stateid].
static int	allocframenum
	Allocated frames of the cache.
static int	allocblock
	Allocation block size per allocateion of the cache.
static LOGPROB *	last_cache
	Local work are to hold cache list of current time.
static LOGPROB *	maxprobs
	Work area that holds N-best state scores for pseudo state set.
static int	maxn
	Allocated length of above.

---

Detailed Description

Computation of acoustic likelihood in HMM state, with state-level cache.

Author:

Akinobu LEE

Date:

Fri Feb 18 18:45:21 2005

This file defines functions to compute output log probability of HMM state. Several functions are defined for each state type (whether it is on word edge and a part of pseudo HMM), and all of them calls outprob_state() to get the log probability of a HMM state. The outprob_state() will set the needed values to the global variables that begins with "OP_", and call calc_outprob_state(). The calc_outprob_state() is actually a function pointer, and the entity is either calc_tied_mix() for tied-mixture model and calc_mix() for others. (If you use GMS, the entity will be gms_state() instead.)

The state scores will be cached here. The 2-dimension cache array of state and input frame are used to store the computed scores. They will be expanded when needed. Thus the scores will be cached for all input frame because they will also be used in the 2nd pass of recognition process.

When using a tied-mixture model, codebook-level cache will be also done in addition to this state-level cache. See calc_tied_mix.c for details.

Revision

1.1.1.1

Definition in file outprob.c.

---

Function Documentation

boolean outprob_cache_init (   )

Initialize the cache data, should be called once on startup. Returns: TRUE on success, FALSE on failure. Definition at line 81 of file outprob.c. Referenced by outprob_init().

boolean outprob_cache_prepare (   )

Prepare cache for the next input, by clearing the existing cache. Returns: TRUE on success, FALSE on failure. Definition at line 98 of file outprob.c. Referenced by outprob_prepare().

static void outprob_cache_extend (  int  reqframe  )  [static]

Expand the cache to time axis if needed. Parameters: reqframe  [in] required frame length Definition at line 118 of file outprob.c. Referenced by outprob_state().

LOGPROB outprob_state (  int  t, HTK_HMM_State *  stateinfo, HTK_Param *  param )

Compute output probability of a state. Set the needed values to the global variables that begins with "OP_", and call calc_outprob_state(). The calc_outprob_state() is actually a function pointer, and the entity is either calc_tied_mix() for tied-mixture model and calc_mix() for others. (If you use GMS, the entity will be gms_state() instead.) The state-level cache is also consulted here. Returns: output log probability. Definition at line 167 of file outprob.c. Referenced by next_word(), outprob(), outprob_cd_avg(), outprob_cd_max(), outprob_cd_nbest(), outprob_style(), and start_word().

void outprob_cd_nbest_init (  int  num  )

Initialize work area for outprob_cd_nbest(). Parameters: num  [in] number of top states to be calculated. Definition at line 204 of file outprob.c. Referenced by outprob_init().

static LOGPROB outprob_cd_nbest (  int  t, CD_State_Set *  lset, HTK_Param *  param )  [static]

Return average of N-beat outprob for pseudo state set. Parameters: t  [in] input frame lset  [in] pseudo state set param  [in] input parameter data Returns: outprob log probability, average of top N states in lset. Definition at line 220 of file outprob.c. Referenced by outprob_cd().

static LOGPROB outprob_cd_max (  int  t, CD_State_Set *  lset, HTK_Param *  param )  [static]

Return maximum outprob of the pseudo state set. Parameters: t  [in] input frame lset  [in] pseudo state set param  [in] input parameter data Returns: maximum output log probability among states in lset. Definition at line 264 of file outprob.c. Referenced by outprob_cd().

static LOGPROB outprob_cd_avg (  int  t, CD_State_Set *  lset, HTK_Param *  param )  [static]

Return average outprob of the pseudo state set. Parameters: t  [in] input frame lset  [in] pseudo state set param  [in] input parameter data Returns: average output log probability of states in lset. Definition at line 286 of file outprob.c. Referenced by outprob_cd().

LOGPROB outprob_cd (  int  t, CD_State_Set *  lset, HTK_Param *  param )

Compute the log output probability of a pseudo state set. Parameters: t  [in] input frame lset  [in] pseudo state set param  [in] input parameter data Returns: the computed log output probability. Definition at line 312 of file outprob.c. Referenced by next_word(), outprob(), outprob_style(), and start_word().

LOGPROB outprob (  int  t, HMM_STATE *  hmmstate, HTK_Param *  param )

Top function to compute the output probability of a HMM state. Parameters: t  [in] input frame hmmstate  [in] HMM state param  [in] input parameter data Returns: the computed log output probability. Definition at line 346 of file outprob.c. Referenced by viterbi_segment().

---