计算机工程与应用 ›› 2007, Vol. 43 ›› Issue (35): 233-234.

• 工程与应用 • 上一篇    下一篇

基于EVRC的带噪汉语语音端点检测

姚艳红,宋 刚

  

  1. 山东大学 信息科学与工程学院,济南 250100
  • 收稿日期:1900-01-01 修回日期:1900-01-01 出版日期:2007-12-11 发布日期:2007-12-11
  • 通讯作者: 姚艳红

Endpoint detection of noisy Chinese speech based on EVRC

YAO Yan-hong,SONG Gang   

  1. School of Information Science and Engineering,Shandong University,Ji’nan 250100,China
  • Received:1900-01-01 Revised:1900-01-01 Online:2007-12-11 Published:2007-12-11
  • Contact: YAO Yan-hong

摘要: 提出了一种基于EVRC的端点检测方法。在背景噪声变化的环境下,该方法将语音映射到一个基于心理声学模型的语音矩阵中,通过跟踪噪声,得到的语音矩阵和参数能够适应不同的背景噪声环境。结合汉语语音的特点,使用该方法无需改变门限,即可准确的检测出语音的端点。

关键词: 端点检测, 谱熵, 小波方差, 鲁棒性

Abstract: The endpoint detection method which is put forward in this article is based on EVRC.The speech is mapped to a voice matrix which is based on a model called mind acoustics.We get the sum of voice matrix that can adapt different background noise circumstance through the noise tracking.Connecting with the characteristic of Chinese speech,we can detect the speech endpoint well and truly without changing the threshold.

Key words: endpoint detection, spectrum entropy, wavelet variance, robustness