Abstract: In order to improve Signal Noise Ratio（SNR） of speech signal, an approach of speech enhancement that combines Empirical 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 signals. The noise in three sub signals is adaptively filtered. Three sub signals that are output from the self-adaptive filter are used to reconstruct the enhanced speech signal. Simulation experiments show that the effect of speech enhancement of the proposed approach is better than that of self-adaptive filter.

Key words: Empirical Mode Decomposition, self-adaptive filter, speech enhancement

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。

关键词: 经验模态分解, 自适应滤波, 语音增强