基于EMD和交叉熵的语音端点检测算法  

Speech endpoint detection based on EMD and cross-entropy

作　　者：薛俊韬[1] 翁玉茹 张军[1] XUE Juntao;WENG Yuru;ZHANG Jun(School of Electrical Engineering & Automation, Tianjin University, Tianjin 300072, China)

机构地区：[1]天津大学电气与自动化工程学院,天津300072

出　　处：《计算机工程与应用》2016年第20期149-153,166,共6页Computer Engineering and Applications

基　　金：天津市科技计划项目(No.13ZXCXGX40400;No.13ZXCXGX40500);天津市滨海新区科技计划项目(No.2012-XJR21017)

摘　　要：针对复杂噪声环境下基于经验模态分解(EMD)的端点检测算法准确率低且不能自适应环境问题,提出了一种结合EMD和交叉熵的语音端点检测新算法。算法利用白噪声在各本征模态函数(IMF)中的概率分布是既定的且与幅值无关的EMD分解特性,将衡量语音帧与噪声帧概率分布差异性的交叉熵特征与EMD能量特征相结合,设置自更新检测阈值,实现复杂噪声环境下的语音端点检测。仿真实验证实了该方法在低信噪比以及非平稳噪声情况下具有显著的有效性和优越性。In view of the problem that speech endpoint detection based on Empirical Mode Decomposition(EMD)losesits accuracy and adaptive in adverse environments, this paper proposes a novel speech endpoint detection algorithm basedon EMD and cross-entropy. EMD decomposition characteristic is analyzed that probability distribution of white noise ineach Intrinsic Mode Functions(IMF)is identified and unrelated to noise amplitude. Since probability distribution of whitenoise is different from that of speech signal, cross-entropy is used to reflect the difference of speech-frames andnoise-frames. EMD-energy feature and cross-entropy are complementary so that they are combined to be a comprehensivedetermination for speech endpoint detection. Adaptive threshold is set to adapt to negative environments. It catches thechanges of noise energy and then it is self-updated to improve accuracy in speech endpoint detection. Simulation resultsindicate that it is effective and superior in the presence of low Signal-to-Noise Ratio(SNR)and non-stationary noise.

关 键 词：端点检测 经验模态分解(EMD) 交叉熵 自适应门限 低信噪比 

分 类 号：TN912.3[电子电信—通信与信息系统]