汉语语音听写机技术的研究与实现  

Research and Implementation of the Techniques for Chinese Dictation Machines

作　　者：郑方[1] 牟晓隆[1] 徐明星[1] 武健[1] 宋战江 

机构地区：[1]清华大学计算机科学与技术系语音实验室

出　　处：《软件学报》1999年第4期436-444,共9页Journal of Software

基　　金：国家863高科技项目基金

摘　　要：文章从声学基元和词法树两个方面对连续语音识别和汉语语音听写机中声学层面的搜索策略进行了分析，提出了基于统计知识的帧同步搜索算法和基于词法约束的词搜索树结构，构成了声学层面的双层搜索网络．算法中利用了统计知识，包括声学层面的差分状态驻留信息和特征变化量信息等．实验结果表明，基于知识的搜索策略使连续语音识别的性能提高了３６．６％．文章还介绍了Ｎ－Ｇｒａｍ统计语言模型的修正退化频度估计算法和搜索算法原理．通过对多年研究成果的分析，实现了一个汉语语音听写机的引擎。In this paper, the search strategies in the acoustic layer of the CSR (continuous speech recognition) and the CDM (Chinese dictation machine) are addressed in two aspects, the acoustic recognition unit and the syntaxconstrained word search tree. The SKBFSS (statistical knowledge based frame synchronous search) algorithm and the syntaxconstrained WST (word search tree) structure are proposed, they form the TLSN (twolevel search network) in the acoustic layer. The statistical knowledge used by the algorithm includes differential state dwell distribution, the feature difference sum and so on, which result in an improvement of 36.6% in CSR. The principles of a modified backoff estimation algorithm and the search algorithms for the Ngram based language models are also introduced. Finally, by integrating the authors' techniques, a Chinese dictation machine engine (CDME) is implemented. A speakerindependent CDM text editor named ST97 and a voice command system named CMD97 are established for personal computers (PCs) based on the CDME.

关 键 词：汉语语音听写机 汉语信息处理 语音识别 

分 类 号：TP391[自动化与计算机技术—计算机应用技术] TN912.34[自动化与计算机技术—计算机科学与技术]