摘要： 针对不确定的汉语语音关键词识别系统，为每个关键词分别建立模型是非常烦琐的，采用音节作为次词具有很好的灵活性。通过同传统的关键词识别系统作比较后发现用音节作为次词的系统具有相当高的识别率，其运算效率也非常高。

关键词: 隐含马尔可夫模型, 关键词识别, 次词

Abstract: It is very difficult to establish an HMM modal for each keyword when the keyword is not pre-defined. Choosing syllable as subword can solve this problem. Compared to traditional keyword spotting system, syllable based keyword spotting system has high performance and efficiency.

中图分类号: 

• TP391.4