低信噪比下基于功率谱熵的语音端点检测算法  

Power spectrum entropy based voice activity detection algorithm in low signal-to-noise ratio conditions

作　　者：刘荣[1] 刘珩[1] 

机构地区：[1]中国农业大学信息与电气工程学院,北京100083

出　　处：《计算机工程与应用》2009年第33期122-124,131,共4页Computer Engineering and Applications

摘　　要：为了解决短波通信中语音检测的问题,针对短波语音信噪比低,噪声复杂的特点,对幅度谱熵算法进行了修正,选取功率谱熵作为VAD特征,加入谱熵平滑和hangover设计,研究了基于功率谱熵的语音端点检测算法。实验证明,算法对几种典型的短波语音均有比较理想的性能。To resolve the problem of voice activity detection in radio communication, considering the characteristics of short wave radio speech has low signal-to-noise ratio and complex noise types,this paper presents a modified voice activity detection algorithm based on power spectrum entropy.The design of hangover and smoothness of power spectrum entropy are equipped to the algorithm.Several kinds of typical radio speech are used to evaluate the algorithm.Experimental results show that this algorithm manifests outstanding performance in radio communication environments.

关 键 词：语音信号处理 谱熵 语音端点检测 低信噪比 

分 类 号：TN912.35[电子电信—通信与信息系统]