Julius 4.2
構造体 FSBeam

Work area for the first pass. [詳細]

#include <recog.h>



TOKEN2tlist [2]
 Token space to hold all token entities.
TOKENIDtindex [2]
 Token index corresponding to tlist for sort.
int maxtnum
 Allocated number of tokens (will grow)
int expand_step
 Number of tokens to be increased per expansion.
boolean expanded
 TRUE if the tlist[] and tindex[] has been expanded at last create_token();.
int tnum [2]
 Current number of tokens used in tlist.
int n_start
 Start index of in-beam nodes on tindex.
int n_end
 end index of in-beam nodes on tindex
int tl
 Current work area id (0 or 1, swapped for each frame)
int tn
 Next work area id (0 or 1, swapped for each frame)
 Active token list that holds currently assigned tokens for each tree node.
int totalnodenum
 Allocated number of nodes in token.
 Special token for beginning-of-sentence.
boolean nodes_malloced
 Flag to check if tokens already allocated.
LOGPROB lm_weight
 Language score weight (local copy)
LOGPROB lm_penalty
 Word insertion penalty (local copy)
LOGPROB lm_penalty_trans
 Additional insertion penalty for transparent words (local copy)
LOGPROB penalty1
 Word insertion penalty for DFA (local copy)
boolean in_sparea
 TRUE when we are in a pause area now.
int tmp_sparea_start
 Memorize where the current pause area begins.
WORD_ID last_tre_word
 Keep ths max word hypothesis at the end of this segment for as the starting word of the next segment.
boolean first_sparea
 TRUE when we are in the first pause area.
int sp_duration
 Number of current successive sp frame.
char * pausemodelnames
 pause model name string to detect segment
char ** pausemodel
 each pause model name to detect segment
int pausemodelnum
 num of pausemodel


Work area for the first pass.

recog.h114 行で定義されています。