计算机与现代化 ›› 2009, Vol. 1 ›› Issue (10): 169-3.doi: 10.3969/j.issn.1006-2475.2009.10.047

• 语音识别 • 上一篇    下一篇

基于连续HMM语音识别系统的构建与分析

张俊,危韧勇   

  1. 中南大学信息科学与工程学院,湖南 长沙 410075
  • 收稿日期:2008-10-27 修回日期:1900-01-01 出版日期:2009-10-15 发布日期:2009-10-15

Construction and Analysis of Speech Recognition System of Continual HMM

ZHANG Jun,WEI Ren-yong   

  1. Institute of Information Science and Engineering, Centre South University, Changsha 410075, China
  • Received:2008-10-27 Revised:1900-01-01 Online:2009-10-15 Published:2009-10-15

摘要: 介绍实现语音识别系统的一些基本概念,对双门限端点检测过程用流程图的形式做详细介绍。该系统采用连续隐马尔科夫模型(HMM)进行建模,对实现过程遇到的三个基本问题进行分析并给出了解决方案。利用该系统对不同人的语音命令进行识别,准确率可达95%。

关键词: 语音识别, 端点检测, 隐马尔科夫

Abstract: Some basic concepts of speech recognition are introduced. And the course of dual threshold endpoint detection is detailedly introduced by flow chart. Continual HMM was used as model of system. The three basic problems met in the course of realizing system are analyzed and the corresponding solutions are provided. The accuracy can reach to 95% by using system which recognized different voice commands of different people. 

Key words: speech recognition, endpoint detection, HMM