一种基于双元麦克风线性阵的语音增强方法  

A method for speech enhancement based on two-elements microphone linear array

作　　者：王三山[1] 何培宇[1] 段文峰[1] 何悦[1] 

机构地区：[1]四川大学电子信息学院,成都610064

出　　处：《四川大学学报（自然科学版）》2009年第2期407-412,共6页Journal of Sichuan University(Natural Science Edition)

基　　金：国家自然科学基金(60472096)

摘　　要：针对现有单通道语音增强算法及传统波束形成算法的局限性,提出了一种基于双元麦克风线性阵的语音增强方法。首先利用离线设计好的优化权值对输入信号进行加权求和以实现波束形成,然后结合一种新的噪声幅度谱估计方法,采用改进的幅度谱减法进一步增强语音信号。仿真实验表明该方法简单易行并取得了较好的语音增强效果。A method for speech enhancement based on two-elements microphone linear array is presented because of the limitation of existed single channel speech enhancement algorithm and conventional beamforming. Using the optimal weights designed off-line, the beamformer output is first obtained from a weighted sum of the input signals. Then combined with a new noise amplitude spectrum estimator, the speech signal is further enhanced by modified amplitude spectral subtraction. The experiments of simulation reveal the simplicity and practicability of the proposed method as well as its good speech enhancement performance.

关 键 词：语音增强 麦克风阵列 波束形成 谱减法 噪声幅度谱估计 

分 类 号：TN912.35[电子电信—通信与信息系统]