汉语语音识别中声学界标点引导的随机段模型解码算法  

Landmark Guided Segmental Speech Decoding Algorithm for Continuous Mandarin Speech Recognition

作　　者：晁浩[1,2] 杨占磊[2] 刘文举[2] 

机构地区：[1]河南理工大学计算机科学与技术学院,焦作454000 [2]中国科学院自动化研究所模式识别国家重点实验室,北京100190

出　　处：《计算机科学》2013年第10期208-212,共5页Computer Science

基　　金：国家自然科学基金(91120303;90820303;90820011);国家重点基础研究发展计划(973计划)(2004CB318105);国家高技术研究发展计划(863计划)(20060101Z4073;2006AA01Z194)资助

摘　　要：提出了一种随机段模型的解码优化算法。检测出具有语音学意义的界标点,根据这些界标点分析临近语音段的边界信息和声韵母类别信息,最后将这些边界信息和类别信息用于指导随机段模型的搜索过程。实验中,两种类型的界标点能较为准确地被检测出来,并用于指导随机段模型的解码,在"863-test"测试集上进行的汉语连续语音识别实验显示,在正确率只有轻微下降的同时,解码时间下降了12.92%,这表明了将语音学知识引入语音识别系统的有效性。A framework was proposed which attempts to incorporate landmarks into segment based Mandarin speech recognition system. In the method,landmarks provide boundary information and phonetic class information, and the in- formation is used to direct the decoding process. To prove the validity of this method, two kinds of landmarks which can be detected reliably were used to direct the decoding process of a segment model（SM）based Mandarin LVCSR system. Experiments conducted on ＂863-test＂ set show that decoding time can be saved about 12. 92% without obviously de-creasing the recognition accuracy. Thus,potential of the method is demonstrated.

关 键 词：语音识别 随机段模型 解码 界标点 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]