通知公告
-
学术报告通知
来源: 发布日期:2016年01月06日题 目: Coupled Canonical Polyadic Decomposition for Joint Blind Source Separation
报告人: Dr. Gong Xiao-Feng (Dalian University of Technology)
时 间: 2016年1月7日 上午9:30
地 点: 信息科学实验楼(10号教学楼)205会议室
Abstract:
Joint blind source separation (J-BSS) is a data driven technique for multiset data fusion, which has attracted increasing interests in many applications such as biomedical engineering and array processing. In this talk, we address the J-BSS problem from a tensorial perspective. We will show, by the use of statistics, that the multi-set data model in J-BSS can be converted to a specific coupled canonical polyadic decomposition (C-CPD) formulation. We have proposed several algebraic and iterative algorithms for the computation of C-CPD. Simulations have shown that, compared with conventional single-set based CPD, C-CPD has relaxed uniqueness conditions, better estimation accuracy of the factor matrices, and naturally aligned permutations among components from different datasets. Several real-life examples, including speech separation, fetal electrocardiogram (ECG) extraction, multi-subject functional magnetic resonance imaging (fMRI) data analysis, etc., are given to illustrate the interest of using C-CPD based J-BSS in practical applications.
Biography:
Dr. Gong Xiao-Feng earned the B.S. degree (Information engineering) from Beijing Institute of Technology, China in 2003. From 2003 to 2009, he took the combined master and Ph.D programme of Beijing Institute of Technology, and earned the Ph.D degree (Communication and information system) in 2009. He was a lecturer from 2009 to 2012, and has been an associate professor from 2013, both with Dalian University of Technology, Dalian, China. He was a visiting research associate with KU Leuven, Belgium. His research interests include multilinear algebra and its applications in blind source separation and array processing.