構造体 __node__

第2パスの文仮説 [詳細]

#include <search.h>

__node__のコラボレーション図

Collaboration graph
[凡例]

変数

__node__next
 Link to next hypothesis, used in stack
__node__prev
 Link to previous hypothesis, used in stack
boolean endflag
 TRUE if this is a final sentence result
WORD_ID seq [MAXSEQNUM]
 Word sequence
short seqnum
 Length of seq
LOGPROB score
 Total score (forward+backward, LM+AM)
short bestt
 Best connection frame of last word in word trellis
short estimated_next_t
 Estimated next connection time frame (= beginning of last word on word trellis): next word hypothesis will be looked up near this frame on word trellis
LOGPROBg
 Current forward viterbi score in each frame
LOGPROB final_g
 Extra forward score on end of frame for multipath mode
int state
 (dfa) Current DFA state ID
TRELLIS_ATOMtre
 Trellis word of last word
LOGPROBg_prev
 Viterbi score back to last 1 phoneme
HMM_Logicallast_ph
 Last applied triphone
boolean last_ph_sp_attached
 Last phone which the inter-word sp has been attached for multipath mode
LOGPROB lscore
 N-gram score of last word (will be used for 1-phoneme backscan and graph output, always 0 for dfa
LOGPROB totallscore
 (n-gram) Accumulated language score (LM only)
LOGPROB cmscore [MAXSEQNUM]
 Confidence score of each word
WordGraphprevgraph
 Graph word corresponding to the last word
WordGraphlastcontext
 Graph word of next previous word
LOGPROB tail_g_score
 forward g score for later score adjustment
__recogprocess__region

説明

第2パスの文仮説

search.h88 行で定義されています。


この構造体の説明は次のファイルから生成されました:
Juliusに対してThu Jul 23 12:17:21 2009に生成されました。  doxygen 1.5.1