作者投稿和查稿 主编审稿 专家审稿 编委审稿 远程编辑

计算机工程 ›› 2012, Vol. 38 ›› Issue (19): 170-174. doi: 10.3969/j.issn.1000-3428.2012.19.044

• 人工智能及识别技术 • 上一篇    下一篇

基于优选信息熵的语音端点检测方法

张 敏,曾晓辉   

  1. (成都信息工程学院通信工程学院,成都 610225)
  • 收稿日期:2011-11-28 出版日期:2012-10-05 发布日期:2012-09-29
  • 作者简介:张 敏(1981-),女,讲师,主研方向:智能信号处理,目标跟踪;曾晓辉,博士研究生

Speech Endpoint Detection Method Based on Optimum Information Entropy

ZHANG Min, ZENG Xiao-hui   

  1. (College of Communication Engineering, Chengdu University of Information Technology, Chengdu 610225, China)
  • Received:2011-11-28 Online:2012-10-05 Published:2012-09-29

摘要: 为提高噪声环境中语音端点检测的准确率,提出一种基于信息熵的检测方法。将分帧语音信号按照不同阶数重新量化,选择其中波动范围大的信息熵作为该信号的优选信息熵,通过多次仿真实验确定较优门限,设计状态机对多段带噪语音进行端点检测。实验结果表明,该方法具有较好的抗噪声性能,在同等环境中的检测误判率较低。

关键词: 端点检测, 波动范围, 信息熵, 门限, 状态机, 误判率

Abstract: To enhance the accuracy of endpoint detection in noisy environment, a detection method based on optimum information entropy is proposed. According to the method, framed speeches are re-quantized with different groups of quantization level, and the group which has greatest range of entropy is chosen to calculate the optimum information entropy of the noisy speech. Thresholds are set by simulations and a state machine is employed to detect the endpoints of noisy speeches. Experimental results show that the method has better noise immunity and lower misjudgment rate.

Key words: endpoint detection, variation range, information entropy, threshold, state machine, misjudgment rate

中图分类号: