多窗谱减法和EMD改进的新型能熵比端点检测  

Multitaper spectrum and EMD improved new energy entropy ratio endpoint detection

作　　者：朱俞清 章小兵 黄镇坤 ZHU Yuqing;ZHANG Xiaobing;HUANG Zhenkun(School of Electrical and Information Engineering,Anhui University of Technology, Ma Anshan 24300, China)

机构地区：[1]安徽工业大学电气与信息工程学院

出　　处：《电声技术》2019年第9期17-21,共5页Audio Engineering

摘　　要：在语音信号中测出语音的端点是极其重要的。在传统的端点检测方法有能零比,能熵比,频带方差法等。这些方法在信噪比较高的环境下可以准确的检测出语音的端点,但在信噪比相对较低的环境下检测的准确率就很低。本文提出了一种在对语音进行多窗谱减法降噪以后,在进行EMD改进的新型能零比特征参数语音端点检测的算法。这种算法通过对语音两个端点检测特征参数Teager能量和基本谱熵进行研究并提出新的语音参数,即为EMD改进的能熵比值来实现语音的端点检测。此算法仿真实验表示,与传统的能熵比端点检测法相比,该算法在不同低信噪比情况下有较高的端点检测正确率。In speech signal measured voice endpoint is extremely important,.In the traditional endpoint detection method can have energy-zero energy-entropy ratio, frequency band variance method, etc. These methods can accurately detect the voice endpoint in the environment of high signal noise ratio (SNR), but in low SNR environment detection accuracy is very low. in this paper to proposed multitaper spectrum subtraction speech after the noise reduction, in a new type of can carry on the improvement of the EMD zero than feature parameters of speech endpoint detection algorithm This algorithm based on the speech endpoint detection two characteristic parameters Teager energy and basic spectrum entropy study and put forward a new voice parameters, namely for the EMD improved new energy entropy ratio to achieve speech endpoint detection the algorithm simulation results, said compared with the traditional entropy than endpoint to test, the algorithm under the condition of low signal-to-noise ratio have higher endpoint detection accuracy.

关 键 词：端点检测 多窗谱减法 Teager能量 能熵比 EMD改进的能熵比 

分 类 号：TN912.3[电子电信—通信与信息系统]