汉语语音的非线性动力学特征及其降噪应用

基金项目:

国家杰出青年科学基金和国家自然科学基金资助项目

详细信息

Nonlinear characteristics of Chinese speech and its application in noise reduction

State Key Lab of Modern Acoustics and Institute of Acoustics, Nanjing University Nanjing 210093

摘要: 分析了汉语语音的相关维、最小嵌入维数以及重构相图。分析结果表明汉语语音具有混沌特征。根据这些非线性特征可以有效区分汉语中的浊音、清音和随机噪声,从而可以用于语音降噪。介绍了本地投影法混沌语音降噪的原理与算法,并利用该算法对一些典型的元音和辅音进行降噪,获得了较好的降噪效果。

Abstract: Some invariants of nonlinear dynamics of Chinese speech are analyzed. The results reveal the chaotic characters in Chinese speech and the difference among voiced, unvoiced sound and random noise, which verify the feasibility of denoising Chinese speech using noise reduction algorithms for chaotic data. The results of denoising experiments show good effects on voiced and unvoiced sounds.