语音识别中的搜索策略研究

在语音识别应用中,如何提高识别的效率是一个重要的研究方向,尤其在大词表的识别中,巨大的搜索空间带来了很大的计算代价。先介绍了语音识别系统的三个组成部分即特征提取、声音模型及语言模型;然后重点讨论了语音识别中几种常用的搜索策略,最后指出了搜索策略进一步研究的方向。

Research on Search Strategy in Speech Recognition

In application of speech recognition,how to improve its efficiency is an important orientation of research especially for large vocabulary speech recognition,because its huge search space results in vast computational cost.First,we introduces the three fundamental components of a typical speech recognition-feature extraction,acoustic model and language model,then we mainly discusses some commonly used search strategy.Finally,we points out the direction of search strategy for future research.