佤语语音语料端点检测算法  

A new endpoint detection algorithm for speech corpus of the Wa language

作　　者：和丽华 江涛[1] 潘文林[1] 杨建香 解雪琴 王璐 余彩裙 HE Li-hua;JIANG Tao;PAN Wen-lin;YANG Jian-xiang;XIE Xue-qin;WANG Lu;YU Cai-qun(School of Mathematics and Computer Science,Yunnan Minzu University,Kunming 650500,China)

机构地区：[1]云南民族大学数学与计算机科学学院,云南昆明650500

出　　处：《云南民族大学学报（自然科学版）》2019年第2期186-190,196,共6页Journal of Yunnan Minzu University:Natural Sciences Edition

基　　金：国家自然科学基金(61363022);云南民族大学研究生创新基金(2018YJCXS233)

摘　　要：端点检测是语音信号处理的过程中非常重要的一个环节,其准确性直接影响语音信号处理的速度和结果.特别是在实际应用中因信噪比较低,使得某些高信噪比下性能好的端点检测算法准确率也比较低.为了提高在低信噪比的环境下佤语语音端点检测的准确率,本文使用了一种基于多窗谱估计谱减法和能熵比法的语音端点检测复合算法.该算法首先利用多窗谱估计谱减法去除语音的背景噪音以提高信噪比;其次再对去噪后的语音使用能熵比算法进行端点检测;最后借助Matlab工具对佤语语音进行仿真实验.仿真结果表明:对于低信噪比的环境下的佤语语音,本文使用的基于多窗谱估计谱减法和能熵比法复合算法同常规能熵比算法相比,端点检测的准确率提高了34%.Endpoint detection is a very important part of the process of speech signal processing,and its accuracy directly affects the speed and result of speech signal processing. Especially in practical applications,due to the low signal-to-noise ratio,the accuracy of the endpoint detection algorithm with good performance at some high SNR is relatively low. In order to improve the accuracy of the endpoint detection of the speech sounds of the Wa language in the environment of a low signal-to-noise ratio,this paper puts forward a new composite algorithm based on the multi-taper spectral estimation of spectral subtraction and energy-entropy ratio,and applies it to the endpoint detection of the speech sounds of the Wa language. Firstly,the background noise is removed by the multi-taper spectral estimation of spectral subtraction to improve the signal-to-noise ratio of the speech sounds of the Wa language. Then it uses the algorithm of the energy-entropy ratio to detect the endpoint detection of the denoised speech sounds of the Wa language. Finally,it uses the Matlab tool to simulate the speech sounds of the Wa language. The simulation results show that the accuracy of the endpoint detection by this algorithm is improved up to 34% compared with the conventional energy-entropy ratio algorithm in a low SNR environment.

关 键 词：佤语语音 多窗谱估计谱减法 能熵比 端点检测 

分 类 号：TN912.34[电子电信—通信与信息系统]