一种针对欠定时延混合模型的混合矩阵估计算法  

A mixing matrix estimation algorithm for the under-determined time-delayed hybrid model

作　　者：聂伟 赵耀东 曹旭源 张圣鹋 白尕太 NIE Wei;ZHAO Yao-dong;CAO Xu-yuan;ZHANG Sheng-miao;BAI Ga-tai(Science and Technology on Electronic Information Control Laboratory,Chengdu 610036,China)

机构地区：[1]电子信息控制重点实验室

出　　处：《信息技术》2020年第2期89-92,102,共5页Information Technology

摘　　要：文中主要研究了欠定时延混合模型下盲源分离的混合矩阵估计问题,提出了一种针对复数混合矩阵估计的新算法。文中首先介绍了欠定时延混合模型;其次通过利用不同观测信号在时频点处模值的内在关系来检测单源点并筛选出用于聚类的单源点;最终采用块聚类的方法估计出混合矩阵。公式推导和仿真分析都证明该算法可以估计出欠定时延混合模型下的混合矩阵。与其它算法相比,该算法具有更好的估计性能。The mixing matrix estimation problem of blind source separation for the under-determined time-delayed hybrid model is researched in this paper and a new algorithm for complex mixing matrix estimation is proposed.First,the under-determined time-delayed hybrid model is introduced in the paper.Then,a method is proposed to detect and select the single source points for clustering by utilizing the internal relationship of different observed signals’modulus at time-frequency points.Finally,the block clustering is used to get the mixing matrix.The formula derivation and simulation results both prove that the proposed algorithm can estimate the mixing matrix of the under-determined time-delayed hybrid model.Compared with other algorithms,the proposed algorithm in this paper owns better estimation performance.

关 键 词：欠定时延混合模型 盲源分离 单源点 混合矩阵 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]