EURASIP - Workshop on Sparsity and Compressive SensingWelcome!
Signal processing methods relying on "sparse representations" have proven useful in compression, de-noising, classification, and inverse problems in imaging, acoustics/speech, and communications. Sparse estimation (or sparse recovery) is playing an increasingly important role in the statistics and machine learning communities. Furthermore the developments in sparse signal processing have also deepened our understanding of sampling, coding, spectral estimation, array processing, component analysis, multipath channel estimation. This workshop will consist of a tutorial session followed by presentations on novel results of compressive sensing and sparse signal processing techniques.
Information
Venue : Silence Beach Resort, Side Antalya
Date : Saturday 11 April, 2009
The workshop is co-located with the 17th Conference on Signal Processing, Communications and their Applications (SIU)2009.
Program Organizers
Bülent Sankur, Boğaziçi University
Müjdat Çetin, Sabancı University
Workshop Program
| Speakers | Title | Time |
| 10.15-11.00 | ||
| 11.15-12.00 | ||
| Lunch Break | ||
| 13.15-14.00 | ||
| 14.15-15.00 | ||
| 15.30-16.15 | ||
| Panel on "Sparsity Techniques in Signal Processing"
Güneş Kurt: Application of Basis Selection Algorithms to Communication Problems Ali Cafer Gürbüz: A Compressive Sensing Data Acquisition and Imaging Method for Ground Penetrating Radars |
||
Lecturers
Volkan Cevher, Research Scientist, Rice University, USA
Website: http://www.ece.rice.edu/~vc3/
Short Bio: Volkan Cevher received his B.S. degree in Electrical Engineering from Bilkent University, Ankara, Turkey in 1999 as a valedictorian. During summer of 2003, he was employed by Schlumberger Doll Research. In Fall 2004, he was awarded the Center for Signal and Image Processing Outstanding Research Award for excellence in signal processing research. He received his Ph.D. degree in Electrical Engineering from Georgia Institute of Technology in May 2005 under the supervision of Dr. James H. McClellan. After his Ph.D., he worked with Dr. Rama Chellappa at the University of Maryland for two years as a research associate on computer vision problems. He is currently working with Dr. Richard Baraniuk at Rice University as a research scientist on compressive sensing problems. His research interests include sensor network problems, compressive sensing, graphical models, computer vision, Monte-Carlo Markov chain methods, and array signal processing.
Title: Compressive Sensing Theory and Applications (download)
Abstract: Compressive sensing (CS) is a new approach to simultaneous sensing and compression that enables a potentially large reduction in the sampling, communication and computation costs at a sensor for signals having a sparse representation in some basis. The CS measurement process is nonadaptive, and the recovery process is nonlinear, for which a variety of algorithms have been proposed.
The goal of this tutorial is to expose the CS theory to a wide audience in academia and industry who are interested in information processing in sensing systems. The tutorial will present the fundamentals of CS in an approachable manner, aiming to encourage engineers in industry and academia to exploit the new theory in their applications and their research. Although several theoretical results will be presented, the emphasis is on the intuition and the understanding of the theory.
Title: Model-based Compressive Sensing (download)
Abstract: Compressive sensing (CS) is an alternative to Shannon/Nyquist sampling for acquisition of sparse or compressible signals that can be well approximated by just K < < N elements from an N-dimensional basis. Instead of taking periodic samples, we measure inner products with M < N random vectors and then recover the signal via a sparsity-seeking optimization or greedy algorithm. The standard CS theory dictates that robust signal recovery is possible from M=O(K log(N/K)) measurements. The implications are promising for many applications and enable the design of new kinds of analog-to-digital converters, cameras and imaging systems, and sensor networks.
While this represents significant progress from Nyquist-rate sampling, in this talk, I will demonstrate that it is possible to do even better by more fully leveraging concepts from state-of-the-art signal compression and processing algorithms. In many such algorithms, the key ingredient is a more realistic signal model that goes beyond simple sparsity by codifying the inter-dependency structure among the signal coefficients. I will present a new model-based CS theory that parallels the conventional theory and provides concrete guidelines on how to create model-based recovery algorithms with provable performance guarantees. To demonstrate the new theory, I will show how to integrate two relevant signal models --wavelet trees and clustered sparsity-- into two state-of-the-art CS recovery algorithms and prove that they offer robust recovery from just M=O(K) measurements.
Jared Tanner, Department of Mathematics, University of Edinburgh, Scotland, UK
Website: www.maths.ed.ac.uk/~tanner/
Short Bio: Jared Tanner received a Ph.D. in Mathematics from the University of California at Los Angeles in 2002. His research contributions have been honored with the Fox Prize (2003), Monroe Martin Prize (2005), Sloan Fellowship (2007) and Leverhulme Prize (2009). Dr. Tanner holds a Warnock Chair Assistant Professorship in Mathematics at the University of Utah and is a Lecturer of Applied Mathematics at the University of Edinburgh. His research interests include mathematical signal processing, numerical analysis, and approximation theory.
Title: Phase Transitions Phenomenon in Compressed Sensing (download)
Abstract: Compressed Sensing has broken the Nyquist barrier, but what is the sampling theorems for CS? Reconstruction algorithms typically exhibit a zeroth-order phase transition phenomenon for large problem sizes, where there is a domain of problem sizes for which successful recovery occurs with overwhelming probability, and there is a domain of problem sizes for which recovery failure occurs with overwhelming probability. These phase transitions serve as sampling theorems for CS. The mathematics underlying this phenomenon will be outlined for L1 regularization and non-negative feasibility point regions. Both instances employ a large deviation analysis of the associated geometric probability event. These results give precise if and only if conditions on the number of samples needed in Compressed Sensing applications. Lower bounds on the phase transitions implied by the Restricted Isometry Property for Gaussian random matrices will also be presented for the following algorithms: Lq-regularization for q between zero and one, CoSaMP, Subspace Pursuit, and Iterated Hard Thresholding.
Müjdat Çetin, Faculty of Engineering and Natural Sciences
Sabancı University, Istanbul Turkey
Website: http://people.sabanciuniv.edu/mcetin
Short Bio: Mujdat Cetin received the B.S. degree from Bogazici University, Istanbul, Turkey, in 1993; the M.S. degree from the University of Salford, Manchester, U.K., in 1995; and the Ph.D. degree from Boston University, Boston, MA, USA, in 2001, in electrical engineering. From 2001 to 2005, he was with the Laboratory for Information and Decision Systems, M.I.T., Cambridge, MA, USA. Since September 2005 he has been an Assistant Professor at Sabanci University, Istanbul, Turkey. He has served in various organizational capacities, including special session organizer, session chair, and technical program committee member for a number of conferences including the IEEE International Conference on Acoustics, Speech, and Signal Processing; the SPIE Conference on Algorithms for Synthetic Aperture Radar Imagery; the IEEE Statistical Signal Processing Workshop; the IEEE International Conference on Image Processing; and the EURASIP European Signal Processing Conference. He has also served as the Technical Program Co-chair for the 2006 IEEE Turkish Conference on Signal Processing, Communications and their Applications. His research interests include statistical signal and image processing, inverse problems, computer vision, data fusion, wireless sensor networks, biomedical information processing, radar imaging, brain computer interfaces, machine learning, and sensor array signal processing. Dr. Cetin has received a number of awards including the 2008 Turkish Academy of Sciences Young Scientist Award, and the 2007 Elsevier Signal Processing Journal Best Paper Award.
Title: Sparsity-Driven Radar Imaging (download)
Abstract: We present some of our recent work on complex-valued coherent image reconstruction, in particular on synthetic aperture radar (SAR) imaging. One of the motivations for our work has been the increased interest in using reconstructed images in automated decision-making tasks. The success of such tasks (e.g. target recognition in the case of radar) depends on how well the computed images exhibit certain features of the underlying scene. Based on the observation that typical underlying scenes usually exhibit sparsity in terms of such features, we have developed image formation algorithms which formulate the imaging problem as a sparse signal representation problem. We present the basic form of our approach, as well as extensions to a number of scenarios motivated by emerging applications including sparse-aperture imaging, passive radar imaging, wide-angle imaging, and imaging in the case of frequency-band omissions. We also mention some of our work on imaging in the case of model errors, as well as our work on automatic hyperparameter choice. Finally we also briefly discuss some of our more recent work exploring the question of what existing compressed sensing theory says about various radar imaging scenarios. During the course of this discussion we provide some pointers to our work on equivalence of l0 and lp regularization. If time permits, we will also briefly mention our related work on ultrasound imaging and acoustic array processing.
Pierre Vandergheynst, Electrical Engineering Institute Ecole Polytechnique Fédérale de Lausanne, Switzerland
Website: http://ltspc89.epfl.ch/~vandergh/
Short Bio: Pierre Vandergheynst received the M.S. degree in physics and the Ph.D. degree in mathematicalphysics from the Université catholique de Louvain, Louvain-la-Neuve, Belgium, in 1995 and 1998, respectively. From 1998 to 2001, he was a Postdoctoral Researcher with the Signal Processing Laboratory, Swiss Federal Institute of Technology (EPFL), Lausanne, Switzerland. He was Assistant Professor at EPFL (2002-2007), where he is now an Associate Professor. His research focuses on harmonic analysis, sparse approximations and mathematical image processing with applications to higher dimensional, complex data processing. He was co-Editor-in-Chief of Signal Processing (2002-2006) and is Associate Editor of the IEEE Transactions on Signal Processing (2007-present). He has been on the Technical Committee of various conferences and was Co-General Chairman of the EUSIPCO 2008 conference. Pierre Vandergheynst is the author or co-author of more than 50 journal papers, one monograph and several book chapters. He's a senior member of the IEEE, a laureate of the Apple ARTS award and holds seven patents.
Title: Two Variations on Compressed Sensing Architectures (download)
Abstract: Compressed sensing is a new and rapidly developing field which states, roughly, that one can sample signals having a sparse representation with a number of samples proportional to the sparsity. In many situations, this would result in sampling below the Nyquist rate. Although the theoretical results are flourishing, it is also particularly interesting to understand how hardware developments will be impacted by the new theory.
In this talk we will address recent practical developments in compressed sensing, in particular geared toward hardware implementation. First we consider the problem of sampling far below the Nyquist rate signals that are sparse linear superpositions of shifts of a known, potentially wide-band, pulse. This signal model is key for applications such as Ultra Wide Band (UWB) communications or neural signal processing. Following the recently proposed Compressed Sensing methodology, we study several acquisition strategies and show that the approximations recovered via minimization are greatly enhanced if one uses Spread Spectrum modulation prior to applying random Fourier measurements. We complement our experiments with a discussion of possible hardware implementation of our technique.
Then we present a CMOS imager with built-in capability to perform Compressed Sensing. The adopted sensing strategy is the random Convolution due to J. Romberg. It is achieved by a shift register set in a pseudo-random configuration. It acts as a convolutive filter on the imager focal plane, the current issued from each CMOS pixel undergoing a pseudo-random redirection controlled by each component of the filter sequence. A pseudo-random triggering of the ADC reading is finally applied to complete the acquisition model. The feasibility of the imager and its robustness under noise and non-linearities have been confirmed by computer simulations, as well as the reconstruction tools supporting the Compressed Sensing theory.
Güneş Kurt, Turkcell R&D Applied Research and Technology Group
Istanbul Turkey
Website:
Short Bio: Dr. Gunes Kurt received the B.S. degree with high honors in electronics and electrical engineering from Bogazici University, Istanbul, Turkey, in 2000. She received her M.A.Sc. and Ph.D. degrees both in electrical engineering from the University of Ottawa, Ontario, Canada in 2002 and 2006, respectively. From 2000 to 2005 she was employed as a Research Assistant at CASP Group, University of Ottawa, where she worked on channel coding, sparse signal representations and their applications to communication problems. She was with TenXc Wireless between 2006 and 2007 as an Algorithmic Engineer, where she worked on location estimation via smart antenna systems. She was with Edgewater Computer Systems Inc. from 2006 to 2008 where she was working on high-bandwidth computer networking in aircraft communication systems. Currently she's with Turkcell R&D Applied Research and Technology group. Her research interests include communications theory, heuristic algorithms, network planning and management. She is the author of 10 journal papers, over 20 conference papers and several patent applications.
Title: Application of Basis Selection Algorithms to Communication Problems (download)
Abstract: The basis selection (BS) problem can be viewed as finding a sparse solution to a linear system of equations. More precisely, if we form a matrix A from the columns of the dictionary D , A = [a1 , a2 , . . . , an ], the problem can be stated as finding a solution, with m non-zero entries with m < n. Sequential basis selection algorithms such as Matching Pursuit can be applied to various communication problems such as signal detection, channel estimation,multi user detection in CDMA systems, and angle of arrival detection for smart antennas. In this talk, we will review application examples to these problems. Previous work shows that application of BS algorithms provide solutions with fast convergence and low number of samples. Comparisons with other existing algorithms such as MUSIC and ESPRIT will be presented.
Ali Cafer Gürbüz, TOBB University of Economics and Technology, Dept. of Electrical-Electronics Engineering, Ankara Turkey
Website: http://alicafergurbuz.com/default.aspx
Short Bio: Ali Cafer Gurbuz received his B.S. in Electrical Engineering from Bilkent University, Ankara, Turkey in 2003 and M.S. and Ph.D. degrees in Electrical and Computer Engineering from the Georgia Institute of Technology, Atlanta/GA, USA in 2005 and 2008, respectively. As a graduate research asssistant (2003-2008) and post-doctoral fellow at Georgia Tech, he conducted research on multi-modal landmine detection systems. He is currently an Assistant Professor at TOBB University of Economics and Technology, Ankara, Turkey in the Department of Electrical and Electronics Engineering. His research interests include compressive sensing, remote sensing and ground penetrating radar applications, fast feature detection techniques and SAR imaging.
Title: A Compressive Sensing Data Acquisition and Imaging Method for Ground Penetrating Radars (download)
Abstract: The theory of compressive sensing (CS) enables the reconstruction of sparse signals from a small set of nonadaptive linear measurements by solving a convex minimization problem. This work involves a novel data acquisition and imaging algorithm for Ground Penetrating Radars (GPR) based on CS by exploiting sparseness in the target space, i.e., a small number of point-like targets. Instead of measuring conventional radar returns and sampling at the Nyquist rate, linear projections of the returned signal with random vectors are taken as measurements. Using simulated and experimental GPR data, it is shown that sparser and sharper target space images can be obtained compared to standard backprojection methods using only a small number of CS measurements. Although the presented results are given for GPR, the developed method is general enough to be applied to a remote sensing application with small modifications.
Menu :
Sponsors :
