A Convex Analysis Framework for Blind Separation of Non-negative Sources

Professor Wing-Kin (Ken) Ma
The Chinese University of Hong Kong

Friday, March 12, 2010 at 2:00pm
Engr IV Maxwell Room 57-124

Abstract
Blind source separation (BSS) is a problem where observations are unknown linear mixtures of source signals and where the task is recover those source signals. Recently there has been a growing interest in a subclass of BSS, known as non-negative BSS (nBSS), motivated by applications such as biomedical imaging, and hyperspectral imaging in remote sensing. This talk presents a new nBSS framework, called convex analysis of mixtures of non-negative sources (CAMNS). Unlike certain existing nBSS methods which are deeply entrenched to the statistical assumption of source independence, CAMNS is deterministic. CAMNS is based on a very special assumption called local dominance, which may be satisfied, or be a good approximation, for source signals exhibiting sparsity or high contrast. Under local dominance and several standard nBSS assumptions, we apply convex analysis to establish a new nBSS criterion, which states that the source signals can be perfectly recovered by identifying the extreme points of an observation-constructed polyhedral set. Effective practical realization methods for achieving the CAMNS criterion are also derived using linear programming. We demonstrate by simulations that CAMNS-based methods can yield significantly better separation performance than other reported nBSS methods.

Biography
Wing-Kin Ma received the B.Eng. degree in electrical and electronic engineering from the University of Portsmouth, Portsmouth, U.K., in 1995 and the M.Phil. and Ph.D. degrees, both in electronic engineering, from the Chinese University of Hong Kong (CUHK), Hong Kong, in 1997 and 2001, respectively. His Ph.D. dissertation was commended to be "of very high quality and well deserved honorary mentioning" by the Faculty of Engineering, CUHK, in 2001. He is currently an Assistant Professor with the Department of Electronic Engineering, CUHK. From 2005 to 2007, he was also an Assistant Professor with the Institute of Communications Engineering, National Tsing Hua University, Taiwan, R.O.C., where he is still holding an adjunct position. Prior to becoming a faculty member, he held various research positions with McMaster University, Canada; CUHK; and the University of Melbourne, Australia. His research interests are in signal processing and communications, with a recent emphasis on MIMO techniques and convex optimization. Dr. Ma currently serves as Associate Editors of the IEEE TRANSACTIONS ON SIGNAL PROCESSING and IEEE SIGNAL PROCESSING LETTERS, and a Guest Editor of IEEE SIGNAL PROCESSING MAGAZINE on the Special Issue titled Convex Optimization for Signal Processing.