EMD与自适应滤波相结合的语音增强法  

Approach of speech enhancement combining EMD with self-adaptive filter

作　　者：侯丽霞[1] 曾以成[1] 熊民权[1] 

机构地区：[1]湘潭大学光电工程系,湖南湘潭411105

出　　处：《计算机工程与应用》2012年第9期104-107,共4页Computer Engineering and Applications

基　　金：湖南省自然科学基金(No.08JJ5031)

摘　　要：为了提高语音信号的信噪比,提出一种经验模态分解与自适应滤波相结合的语音增强法。对带噪语音进行经验模态分解,得到有限个固有模态函数,把所有的固有模态函数按顺序分成三组,将每一组所包含的固有模态函数叠加,得到三个子信号;对三个子信号进行自适应滤波,消除噪声;将降噪后的子信号重构得到增强后的语音。仿真实验表明,所提方法的语音增强效果优于自适应滤波。In order to improve Signal Noise Ratio （SNR） of speech signal, an approach of speech enhancement that combines Empiri- cal Mode Decomposition （EMD） and self-adaptive filter is presented. The noisy speech is decomposed into several Impirical Mode Functions（IMFs） by EMD. All IMFs are separated into three groups and the IMFs of every group are added up to form three sub sig- nals. The noise in three sub signals is adaptively filtered. Three sub signals that are output from the self-adaptive filter are used to recon- struct the enhanced speech signal. Simulation experiments show that the effect of speech enhancement of the proposed approach is bet- ter than that of self-adaptive filter.

关 键 词：经验模态分解 自适应滤波 语音增强 

分 类 号：TN912[电子电信—通信与信息系统]