基于n元语法模型的领域语音指令识别  

Domain Voice Command Recognition Based on N-Gram Model

作　　者：吴应中[1] 朱华勇[1] 李杰[1] 

机构地区：[1]国防科学技术大学机电工程与自动化学院,410073

出　　处：《微计算机信息》2009年第14期234-236,共3页Control & Automation

摘　　要：本文采用一种音素匹配法结合改进了的n元语法模型的方法,对语音指令进行理解。指令按照结构划分类别,通过提取其中关键要素来获取指令语义。音素匹配法可以提取指令中部分或全部要素;对于指令中未识别的要素则根据已正确识别的部分,采用改进的n元语法模型进行推测。实验表明该方法取得了较好的效果,并在原有的基础上有一定的提高。This paper use a method which combines phoneme-match and n-gram model, to understand the voice commands. The commands are classified according to the struct, and to have the means through pick-up the key factors. Phoneme-match method can pick-up all or part of the key factors; for the unrecognized factors in the commands, we can adopt the ameliorative n-gram model and according to those already recognized factors to speculate them. The experiment shows the method obtain a good effect, and take a progress than before.

关 键 词：自然语言理解 音素匹配 n元语法模型 指令理解 

分 类 号：TP391.9[自动化与计算机技术—计算机应用技术]