2008年清华大学信息技术研究院系列学术报告9

Title(1): Convex Analysis Based Unmixing Algorithms for Hyperspectral Imaging

Title(2): Blind ML Detection of Orthogonal Space-Time Block Codes:

Efficient Implementations, Identifiability, and Code Constructions

Speaker(1): Chong-Yung Chi

Professor, Institute of Communications Engineering, & Department of Electrical Engineering National Tsing Hua University, Hsinchu, Taiwan

Speaker(2): Wing-Kin Ma

Assistant Professor, Department of Electronic Engineering

The Chinese University of Hong Kong

Chong-Yung Chi received the Ph.D. degree in Electrical Engineering from the University of Southern California in 1983. From 1983 to 1988, he was with the Jet Propulsion Laboratory, Pasadena, California. He has been a Professor with the Department of Electrical Engineering since 1989 and the Institute of Communications Engineering (ICE) since 1999 (also the Chairman of ICE for 2002-2005), National Tsing Hua University, Hsinchu, Taiwan. He co-authored a technical book, Blind Equalization and System Identification, published by Springer 2006, and published more than 140 technical (journal and conference) papers. His current research interests include signal processing for wireless communications, and statistical signal processing. Dr. Chi is a senior member of IEEE. He has been a Technical Program Committee member for many IEEE sponsored workshops, symposiums and conferences on signal processing and wireless communications, including co-organizer and general co-chairman of IEEE SPAWC 2001, and Co-Chair of Signal Processing for Communications Symposium, ChinaCOM 2008. He was an Associate Editor of IEEE Trans. Signal Processing (5/2001~4/2006), IEEE Trans. Circuits and Systems II (1/2006-12/2007) and an editor (7/2003~12/2005) and a Guest Editor (2006) of EURASIP Journal on Applied Signal Processing. Currently, he is an Associate Editor for the IEEE Signal Processing Letters, an Associate Editor for the IEEE Trans. Circuits and Systems I, and a member of Editorial Board of EURASIP Signal Processing Journal, and a member of IEEE Signal Processing Committee on Signal Processing Theory and Methods.

Wing-Kin Ma received the B.Eng. (with First Class Honors) 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.He is currently an Assistant Professor in the Department of Electronic Engineering, CUHK. He was with the Department of Electrical Engineering and the Institute of Communications Engineering, National Tsing Hua University, Taiwan, also as an Assistant Professor, from August 2005 to August 2007. He is still holding an adjunct position there. Prior to becoming a faculty, he held various research positions at McMaster University, Canada, CUHK, Hong Kong, 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’s 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 Associate Editor of the IEEE TRANSACTIONS ON SIGNAL PROCESSING.

AbstractⅠ

Hyperspectal unmixing aims at identifying the hidden spectral signatures (or endmembers) and their corresponding proportions (or abundances) from an observed hyperspectral scene. In planetary exploration, hyperspectral unmixing provides a powerful tool for analyzing the composition and mineralogy of the observed planetary surfaces. This talk presents two hyperspectral unmixing algorithms using convex analysis. The first algorithm is based on the convex analysis of mixtures of non-negative sources (CAMNS), a very recently developed non-negative blind source separation framework that has not been tested in hyperspectral unmixing. CAMNS has an appealing provable property that it can identify the abundances perfectly (and subsequently the endmembers) under a popular hyperspectral unmixing assumption, known as the pure pixel assumption. The second algorithm, called the minimum simplex volume algorithm (MSVA), considers a more challenging case where no pure pixel is assumed to be existent. We provide a hyperspectral unmixing formulation where the goal is to find the `best' data-enclosing simplex by minimizing the simplex volume. We then propose a novel cyclic minimization procedure that use linear programs (LPs) to sequentially reduce the simplex volume. Both the CAMNS and MSVA are based on solving LPs, and hence they can be efficiently implemented by using readily available LP solvers. Some Monte Carlo simulations and real data experiments are presented to demonstrate the efficacy of the proposed methods over several existing unmixing methods.

AbstractⅡ

The blind maximum-likelihood (ML) detection (or noncoherent ML detection) of a generic space-time block code (STBC) is usually a challenging implementation problem. This talk focuses on the orthogonal STBCs (OSTBCs), a well known scheme that offers simple code structures, low-complexity coherent ML detection, and the maximal transmit diversity. Our investigation indicates that OSTBCs provide many benefits in the blind scenario: The special OSTBC structures not only lead to a significant simplification to the blind ML problem thereby enabling quasi-optimal implementations, but also provide very attractive blind identifiability conditions potentially. First, we will show that the blind ML problem can be simplified to a Boolean quadratic program (BQP) for BPSK or QPSK constellations. The BQP is still a computationally hard problem, and we propose to handle the problem by semidefinite relaxation (SDR), a convex optimization based method that leads to a high-precision approximate blind ML implementation with a polynomial-time worst-case complexity. Another popular BQP solver, namely sphere decoding, will be also considered and compared. Extensions to semiblind detection and OSTBC-OFDM will then be discussed. Second, we will present several key results in our identifiability analysis. Essentially, there exist OSTBCs that are identifiable with probability 1 in Gaussian fading channels, and OSTBCs that are 'perfect' in the sense that they are identifiable for any nonzero channel. But our study also indicates that many of the existing OSTBCs do not fall into the above mentioned classes of codes. To bridge this gap, we will suggest a code construction procedure that can convert almost any BPSK/QPSK OSTBC (that may have poor identifiability) to a perfectly identifiable OSTBC.