首页 | 本学科首页   官方微博 | 高级检索  
     

一种基于检测元音的孤立词端点检测算法
引用本文:邝航宇,张军,韦岗.一种基于检测元音的孤立词端点检测算法[J].电声技术,2005(3):40-43,48.
作者姓名:邝航宇  张军  韦岗
作者单位:华南理工大学,电子与通信工程系,广东,广州,510640
摘    要:提出了基于检测元音的端点检测算法。首先检测到语音中的元音的端点。然后利用元音的端点作为参考端点检测出语音真实的端点。将新方法应用在T146数据包在NoiseX-92的5种噪声下的端点检测和识别实验。并和基于能量的端点检测算法比较。2种不同的实验表明,提出的基于检测元音的孤立词端点检测算法可以在不同信噪比下提高端点检测的准确率,并在低信噪比的环境下能明显提高语音识别系统的识别率。

关 键 词:语音识别  端点检测  自相关函数
文章编号:1002-8684(2005)03-0040-04

A Robust Endpoint Detection Based on Vowel Detection in Isolated Word Recognition
KUANG Hang-y,ZHANG Jun,WEI Gang.A Robust Endpoint Detection Based on Vowel Detection in Isolated Word Recognition[J].Audio Engineering,2005(3):40-43,48.
Authors:KUANG Hang-y  ZHANG Jun  WEI Gang
Affiliation:KUANG Hang-y(u),ZHANG Jun,WEI Gang
Abstract:A new algorithm for endpoint detection which based on the different characteristics of speech and noise is presented. The proposed algorithm relies on detecting the endpoints of the vowels. The starting-point of the first vowel and the ending-point of the last voiced vowel are located to be used as reference endpoints for further determination of the actual endpoint. The new algorithm is evaluated by a task on TI46 database and NoiseX-92 that added to original speech at different SNRs. Experimental results show that the performance of endpoint detection and speech recognition can be improved by this new algorithm in various noisy environments.
Keywords:speech recognition  endpoint detection  autocorrelation function  
本文献已被 CNKI 维普 万方数据 等数据库收录!
设为首页 | 免责声明 | 关于勤云 | 加入收藏

Copyright©北京勤云科技发展有限公司  京ICP备09084417号