语音增强技术在关键词检测系统中仿真研究

The Simulation Research on Speech Enhancent Technique in the Keyword Spotting System

描述了一个机票定制系统,在强噪声环境的公共场所针对汉语小词汇量非特定说话人的连续无限制语音流中检测出其中的关键词语音.为增强其关键词语音信号提出建立一种新的基于行为模式分解的语音增强算法.对同一噪声环境下增强关键词语音信号的HMM与未增强关键词语音的HMM进行了比较,结果显示,采用增强语音信号强度的算法在提高关键词的检测率同时,有效地降低了虚报率,系统的整体性能较好,具有一定的可行性与实用性.

In this paper,a plane ticket advance booking system is described,which aims at Keyword speech and non-keyword speech that present to continuous unconditional speech stream for non-special speaker in Chinese small vocabulary.A new speech enhancent arithmetic based on behavior pattern decomposition is offered,in order to enhance the keyword speech.The comparison of keyword HMM based enhancent keyword speech with the keyword speech is made,the results show that the enhancent keyword speech has a great improvement in the probability detection for the keyword and falling the false report,the whole system has a better capability and the practicability properly.