发明名称 |
SPEECH RECOGNIZING METHOD |
摘要 |
PROBLEM TO BE SOLVED: To carry out highly precise speech recognition making good use of environment-dependent phoneme standard patterns. SOLUTION: While a network is generated by utilizing context free grammar having an environment-dependent phoneme as a termination symbol and an LR(land reserve) purser, a standard pattern is determined and while an input speech is matched against standard patterns of HMM(hidden markov model) a one-pass search is advanced; and environment-dependent phoneme standard patterns as used as the standard patterns of HMM and when an environment- dependent phoneme P3 is predicted by the LR purser from a tip node 3 at the time of the extension of the network, tracing from a new node 4 back to the last node is performed to set the environment-dependent phoneme standard patterns to the arc between nodes 3 and 4 by using phoneme strings P1 , P2 , and P3 as a triphone model. |
申请公布号 |
JPH09244687(A) |
申请公布日期 |
1997.09.19 |
申请号 |
JP19960049898 |
申请日期 |
1996.03.07 |
申请人 |
NIPPON TELEGR & TELEPH CORP <NTT> |
发明人 |
YAMADA TOMOKAZU;SAGAYAMA SHIGEKI |
分类号 |
G10L15/10;G10L15/14;G10L15/18;G10L15/28;(IPC1-7):G10L3/00;G10L3/00 |
主分类号 |
G10L15/10 |
代理机构 |
|
代理人 |
|
主权项 |
|
地址 |
|