データ構造 | |
struct | s_cell |
Element of successor word list for LM factoring computation. [詳細] | |
struct | RC_INFO |
State output probability data for head phone of a word. [詳細] | |
struct | LRC_INFO |
State output probability data for 1-phone word. [詳細] | |
union | ACOUSTIC_SPEC |
State output probability container on lexicon tree. [詳細] | |
struct | LM_PROB_CACHE |
LM cache for the 1st pass [詳細] | |
struct | __A_CELL2__ |
Transition arc holding cell [詳細] | |
struct | wchmm_state |
HMM state on tree lexicon. [詳細] | |
struct | WCHMM_WORK |
wchmm-specific work area [詳細] | |
struct | wchmm_info |
Whole lexicon tree structure holding all information. [詳細] | |
マクロ定義 | |
#define | A_CELL2_ALLOC_STEP 4 |
Number of arcs in an arc cell. | |
型定義 | |
typedef s_cell | S_CELL |
Element of successor word list for LM factoring computation. | |
typedef __A_CELL2__ | A_CELL2 |
Transition arc holding cell | |
typedef wchmm_state | WCHMM_STATE |
HMM state on tree lexicon. | |
typedef wchmm_info | WCHMM_INFO |
Whole lexicon tree structure holding all information. | |
列挙型 | |
enum | AS_Style { AS_STATE, AS_LSET, AS_RSET, AS_LRSET } |
ID to indicate which data is in the ACOUSTIC_SPEC container. [詳細] |
このファイルでは,第1パスで用いられる木構造化辞書(あるいは単語連結 HMM (wchmm) とも呼ばれる)の構造体を定義しています. 起動時に,単語辞書の 前単語が並列に並べられ,ツリー上に結合されて木構造化辞書が構築されます. HMMの状態単位で構築され,各状態は,対応するHMM出力確率,ツリー内での遷移先 のリスト,および探索のための様々な情報(言語スコアファクタリングのための successor word list や uni-gram 最大値,単語始終端マーカー,音素開始 マーカーなど)を含みます.
wchmm.h で定義されています。
enum AS_Style |
ID to indicate which data is in the ACOUSTIC_SPEC container.