一种基于小波的改进型带噪话音激活检测算法  

A modified effective noisy voice activity detection based on wavelets

作　　者：李强[1] 黄冰[1] 王苗[1] 

机构地区：[1]桂林电子科技大学 信息与通信学院,广西桂林541004

出　　处：《微型机与应用》2010年第8期72-75,共4页Microcomputer & Its Applications

基　　金：广西科学基金(0640167)

摘　　要：提出了一种基于小波的改进型话音激活检测方法。在介绍小波变换、增强型变速率语音编码(EVRC)算法中的语音增强模块原理的基础上,着重讲述了利用小波对语音信号进行分解重构,利用语音增强模块中的判别方法对语音端点进行检测,依据判决更新对检测结果进行修正。通过对不同信噪比下的语音信号进行端点检测,给出了仿真结果及语音帧检测率,证明了该算法的有效性和可靠性。In this paper, a modified voice activity detection based on wavelets is proposed. Firstly, it describes the principles of wavelets transform and the speech enhancement of EVRC. Then, the decomposition and reconstruction of speech signal used sym5 wavelet is emphasized and combined to speech enhancement model. According to the update mechanism of pre-judgments, an amendment to the results will be executed at the end of the algorithm. At last, speech signal is detected by this algorithm to get the result of simulation under different signal-to-noise ratios. And the result indicate the effectiveness of this algorithm.

关 键 词：小波变换 语音增强 端点检测 低信噪比 

分 类 号：TP301.6[自动化与计算机技术—计算机系统结构]