Monaural noisy speech separation combining sparse non-negative matrix factorization and deep attractor network  

作　　者：GE Wanying ZHANG Tianqi FAN Congcong ZHANG Tian 

机构地区：[1]School of Communication and Information Engineering,Chongqing University of Posts and Telecommunications,Chongqing 400065

出　　处：《Chinese Journal of Acoustics》2021年第2期266-280,共15页声学学报（英文版）

基　　金：supported by the National Natural Science Foundation of China(61671095,61702065,61701067,61771085);the Project of Key Laboratory of Signal and Information Processing of Chongqing(CSTC2009CA2003);Chongqing Graduate Research and Innovation Project(CYS17219);the Research Project of Chongqing Educational Commission(KJ1600427,KJ1600429)。

摘　　要：The performance of the monaural speech separation method is limited when the speech mixture is disordered by background noise.To obtain the enhanced separated speech from the noisy mixture,a monaural noisy speech separation method combining sparse nonnegative matrix factorization(SNMF)and deep attractor network(DANet)is proposed.This method firstly decomposes the noisy mixture into coefficients of speech and noise respectively.Then the speech coefficient is projected to a high-dimensional embedding space and a DANet is trained to force the embeddings to move to different clusters.The attractor points are used to separate the speech coefficients by masking method,and finally the enhanced separated speeches are reconstructed by the speech basis and their corresponding coefficients.Experimental results in various background noise environments show that the proposed algorithm effectively suppress the noises without decreasing the quality of reconstructed speech by comparison with different baseline methods.

关 键 词：ATTRACTOR SEPARATION FACTORIZATION 

分 类 号：TN912.3[电子电信—通信与信息系统]