Edinburgh International Lecture Series "Advanced Topics in Signal Processing and Communications"


Joint Institute for Signal and Image Processing / University of Edinburgh and Heriot-Watt University


Organiser: Norbert Goertz, Institute for Digital Communications, The School of Engineering and Electronics, The University of Edinburgh

Sponsors: NEWCOM, The Royal Academy of Engineering, Vodafone Group Foundation, The School of Engineering and Electronics





26 March 2007
Prof. Li Ping
City University of Hong Kong

"Interleave-Division Multiple-Access (IDMA), Superposition Coded Modulation (SCM) and OFDM-IDMA"
Video Conference Room, 13:00-14:00 (Slides)

Abstract:
This talk provides an overview of interleave-division multiple-access (IDMA) that employs interleaving as the only mechanism to distinguish users. IDMA possesses many desired features for future wireless systems that are difficult to fulfill simultaneously with other alternatives. These include

The basic IDMA principles can be directly extended from binary modulation to multi-ary modulation. This leads to the so-called superposition coded modulation (SCM) scheme in which all the bandwidth resource is allocated one or a few users. IDMA can also be combined with OFDM. Such combinations provide a solution to both MAI and ISI problems. Interestingly, it also provides a solution to the peak-to-average-power-ratio (PAPR) problem in OFDM. The clipping distortion for peak power reduction can be effectively minimized using a sub-optimal soft compensation technique. Several distinguished features of the new scheme are explained, such as

Comparisons with other coded modulation schemes (in particular, BICM) will be provided.


and

"Multi-User Gain, Maximum Eigenmode Beamforming and IDMA"

Video Conference Room, 14:30-15:30 (Slides)

Abstract:
We consider multi-user MIMO systems with rate constraints. We show that significant performance improvements, quantified by multi-user gain (MUG), can be achieved by a low-cost maximum eigenmode beamforming (MEB) strategy. This technique avoids complicated correlation matrix optimization and water-filling operations at the transmitter (as required by the optimal approach). It provides a simple yet nearly optimal approach to exploit the advantages offered by MIMO. We show that IDMA is an efficient method to realize MUG based on the MEB strategy.



10 October 2007
Prof. Daniel J. Costello, Jr.
University of Notre Dame, Indiana, USA

"The Genesis of Coding Theory"
AGB Seminar Room, 12:00-13:00

Abstract:
This talk gives an historical overview of the theory of channel coding dating back to the work of Shannon in 1948. The important advances in channel coding in each 10 year period since 1948 are viewed from a common perspective: the power and bandwidth efficiencies needed to achieve certain levels of error performance. The most important contributions of the last half century are highlighted, including Hamming codes, Reed-Solomon codes, convolutional codes, soft decision decoding, trellis coded modulation, concatenated codes, turbo codes, low-density parity-check codes, and iterative decoding. Finally, areas of potential future research in channel coding are briefly discussed.

and

"LDPC Convolutional Codes: What are they?  How do they work?  Are they any good?"

AGB Seminar Room, 15:00-16:30

Abstract:
Low-density parity-check (LDPC) convolutional codes have been shown to be capable of achieving the same capacity-approaching performance as LDPC block codes with iterative message-passing decoding.  In this paper, we define two distinct classes of LDPC convolutional codes: time-invariant and time-varying, and connections with quasi-cyclic LDPC block codes are discussed.  Then encoding and decoding procedures are reviewed.  Two iterative message-passing decoders are presented:  a one-shot decoder that treats a terminated LDPC convolutional code as a big block code and a high-speed pipeline decoder that takes advantage of the unique structural properties of convolutional codes and can be used for continuous (non-terminated) transmission.  VLSI implementation requirements for these decoders  are also discussed.  We then compare several aspects of LDPC convolutional codes with LDPC block codes, including encoding complexity, decoding complexity, decoder storage requirements, decoding delay, and error performance.  Finally, a pseudo-codeword analysis is used to investigate the behavior of iterative message-passing decoding for LDPC convolutional codes. In particular, we show that LDPC convolutional codes have better convergence properties than LDPC block codes in the low-to-moderate SNR region due to an improved pseudo-codeword weight spectrum.



12 June 2007
Dr. Marcus Greferath 

University College Dublin / Claude Shannon Institute, Dublin, Ireland

"A class of low-density parity-check codes derived from inversive spaces"

Abstract:
LDPC codes have been attracting attention over the recent decade. They were (re)discovered soon after the famous TURBO codes, and it can be said that they belong to the oldest codes known, since they were originally described in seminal work by Gallager and in the sixties. This talk will describe a family of LDPC codes that are derived from what are called 0-1-geometries which have found in a geometric structure called inversive space. We will briefly discuss basic properties and show some performance diagrams. These diagrams suggest that these codes might be useful in various applications like general communications as well as data storage.



12 June 2007
Dr. Mark Flanagan 

University College Dublin / Claude Shannon Institute, Dublin, Ireland

"On Constructions of girth-8 LDPC Codes from Finite Euclidean Geometries and Finite Multidimensional Lattices"

Abstract:
We demonstrate a construction technique for low-density parity check (LDPC) codes based on m-dimensional finite Euclidean geometries. These codes are shown to be regular Gallager codes with Tanner graphs of girth eight. The minimum distance of these codes is shown to be lower-bounded by 2^m. The codes are also amenable to an efficient partly parallel decoder implementation, which may be used in conjunction with the turbo decoding message passing (TDMP) algorithm for LDPC decoding. It is also shown how results on these codes and their lattices, yielding a code class with highly flexible code length and rate. Finally, simulation results show that both classes of codes have very good error-correcting performance.



27 October 2006
Prof. Sergios Theodoridis
University of Athens, Greece

"Support Vector Machines: A geometric point of view"


Abstract:

Support Vector Machines have been established as one of the major classification and regression tools for Pattern Recognition and Signal Analysis.  Over the last decade a number of theoretical arguments have been developed in order to justify their enhanced performance. The most widely known scenario is to look at them as maximum margin classifiers. Another approach is via learning theory arguments and the structural risk minimization principle, which leads to an optimal trade off between performance and complexity. An alternative path is to look at the cost function, associated with the SVM´s, as a regularized minimizer that asymptotically tends to the Bayesian classifier. A less known viewpoint is the geometric one that leads to the notion of reduced convex hulls. For the non-separable class case, the SVM solution is shown to be equivalent with computing the minimum distance between two reduced versions of the original convex hulls that “encircle” the two classes (for the two class case).In this talk I will focus on the geometric approach and new results will be discussed concerning a) novel, necessary for our case, theorems concerning the structure and properties of the reduced convex hulls (RCH) and b) novel algorithms for computing the minimum distance between the resulting RCH´s. This problem is far from being trivial, since existing algorithms, which compute the minimum distance between convex hulls, rely on their respective extreme points. However, computing the extreme points of a reduced convex hull, as we have shown, is a computationally hard task of a combinatorial nature. A basic projection theorem, that we have shown, will be discussed that bypasses the combinatorial burden of the task and opens the way to employ geometric minimum distance algorithms to the SVM task.  Most important, this theorem “respects” inner products, thus allowing to the well known kernel trick to be easily incorporated into the algorithmic schemes, making them appropriate for the general nonlinear non-separable problem. The derived geometric algorithms are much more efficient compared to the classical and widely used SMO algorithm and its versions. A number of tests with well known test beds have shown that, sometimes,   a gain of an order of magnitude in the number of kernel computations, for similar error rates, can be achieved. Furthermore, the new schemes are closer to our intuitive understanding of an iterative algorithm in simple geometric arguments.  



11 October 2006

Prof. Joachim Hagenauer
Munich University of Technology, Germany


"Information Theory and Genetics"

Abstract:

We view the process between the DNA and the proteins as a communications process. The information transfer between DNA-variations (SNPs) and diseases is measured with Shannon's mutual information using simulated and clinical data (Schizophrenia, Parkinson and Graves autoimmune disease) . We further create distance measures between different DNA sequences derived from mutual information for classification and recognition methods. From classification results we create mammalian and human phylogenetic trees. A communication theory model allows the evaluation of the synchronization process for the DNA to mRNA transcription for E-coli bacteria.



11 May 2006
Prof. Joseph Boutros
Ecole Nationale Supérieure des Télécommunications, Paris, France


"Near outage limit space-time coding for MIMO channels"

Abstract:
We study simple space-time coding techniques for multiple-input multiple-output (MIMO) quasi-static channels (2Tx and 4Tx systems) capable of achieving near outage limit performance.
The core of our space-time code is an h-pi-diagonal state multiplexer that guarantees full diversity and a quasi-optimal coding gain on the MIMO channel. The whole range of word error probability is attained at signal-to-noise ratios extremely close to theoretical limits. In addition, at a fixed signal-to-noise ratio, the word error probability is insensitive to the block length.

and

"Enhanced channel decoding via EM source-channel estimation"


Abstract:

We investigate the joint source-channel estimation and decoding problem. We consider a non-uniform binary source transmitted over a binary-input output-symmetric channel, namely the BEC, BSC and AWGN channels. The source sequence is encoded via systematic and non-systematic low-density parity-check codes. The proposed joint source-channel iterative estimation  technique relies on the Expectation Maximization (EM) algorithm that will be associated to the message passing LDPC decoding for both systematic and non systematic codes. Simulation results confirm the strong improvement in performance over the case in which source a priori information is not considered. Furthermore, within this proposed joint source-channel iterative estimation via Expectation Maximization no loss in error-rate performance is observed with respect to the perfect knowledge case.



20 April 2006
Prof. Mikael Skoglund
Royal Institute of Technology (KTH), Stockholm, Sweden

"Dirty Paper Modulation"

Abstract:
We present a symbol-by-symbol approach to Costa's "dirty paper" precoding problem. Our scheme is based on joint optimization of a modulator (encoder) and detector (decoder), subject to a constraint on the average transmit power. The modulator maps an information symbol (taken from a finite alphabet) and an interference symbol (from the complex field) onto a transmitted constellation point. The detector picks the information symbol (as function of the received symbol) which minimizes the average error probability. We illustrate that the new "dirty paper modulation" scheme outperforms Tomlinson-Harashima precoding, which is a classical suboptimal solution to the known-interference precoding problem. In our simulations, the new approach is also able to perform close to the no-interference bound.



16 January 2006
Prof. Lajos Hanzo
University of Southampton

"Genetics in wireless communications
and 
"Advances in multicarrier systems"



15 November 2005
Dr. Cheng-Xiang Wang
Heriot-Watt University, Edinburgh

"Analog and Digital Mobile Radio Channel Simulators Based on Sum-of-Sinusoids Principles"

Abstract:
A profound understanding and accurate modeling of mobile radio channels are of great significance for the effective design, parameter optimization, and performance evaluation of any mobile communication systems. The sum-of-sinusoids channel modeling approach has widely been accepted as an efficient and flexible method for the simulation of mobile radio channels. This presentation addresses the principles and applications of sum-of-sinusoids channel simulators for both analog (waveform) and digital mobile radio channels.



10 and 11 October 2005 (several talks)
Prof. Ulrich Heute
Christian-Albrechts University, Kiel, Germany

"Integral and Diagnostic Speech-Quality Measurement: State of the Art, Problems, and New Approaches"

Abstract:

The user’s overall impression of a speech signal which has been filtered, transmitted, coded, or processed by some system, can be described in terms of the integral quality. There are well-defined auditory methods to assess integral quality: Mostly, an absolute-category rating in a listening-only situation is used (resulting in the mean-opinion score, MOS). Numerous other methods exist, using different scales or more refined test situations, e.g., conversational tests.

For this integral quality, proposals exist for efficient instrumental, signal-based measurements, yielding MOS estimates, like the standard P.862 (ITU-T, 2001) or the TOSQA model (Berger, 1998). Such estimates have problems with distortions not taken into account during the model development. Furthermore, these estimates do not allow to characterize the quality of speech signals: They might provide the same MOS for two signals which are perceived to sound differently in an auditory test.

A different approach has been well-known that overcomes these problems: Quality attributes can be aimed at, which
- concern detailed and distinct distortion effects,
- thus, allow for a system diagnosis, and
- together form an integral-quality-impression model
- which is able to cope also with future distortions.

Earlier work into that direction has been known to have problems, too. Avoiding some weaknesses and re-defining more suitable attributes, the diagnostic approach is re-visited, including an integral quality measure to be finally derived. New, important types of distortions, like time-varying, bursty data-packet loss in internet transmission, are to be included. First results are reported, and further work is outlined.


"Design of Efficient Digital FIR Filters: A Case Study"

Contents:

1. Introduction                                      
2. Lowpass Tolerance Scheme                          
3. Non-Recursive Filters in 2nd Canonical (Direct) Form
4. Non-Recursive Filters in Cascade Form              
5. Partially Recursive Cascade Structures              
6. Recursive Filters                              
7. Comparison of Expenditure                      
8. Conclusions



26 and 28 July 2005 (several talks)
Prof. Wolfgang Utschick
Munich University of Technology, Germany
"Multiuser multiple input multiple output (MIMO) wireless systems"

Overview:

  1. A tutorial talk about Multiple Input Multiple Output (MIMO) wireless systems which gives an overview of the most exciting topics (tutorial talk).

  2. A tutorial course about the very principles of linear and nonlinear Rx and Tx equalizers with a focus on Tx processing and a strong concentration
    on Tomlinson Harashima Precoding (lecture course)

  3. A couple of accompanying conference talks (from recent conferences) with respect to the topics in 1 and 2 (conference style).




28 June 2005
Dr. Liam O'Carroll
School of Mathematics, University of Edinburgh
"An Introduction to the Mathematical Aspects of Reed-Solomon Codes"

Abstract:
Reed-Solomon codes are arguably the most important family of codes for burst-error correction in channel coding, especially in their use as the outer code in a concatenated code.  Even to discuss them in the binary case (which is the case we consider) involves coming to grips with the arithmetic of finite ('Galois') fields GF(2^m) of 2^m elements.

In this talk, we aim to give a gentle introduction to the mathematical construction and properties of Reed-Solomon codes.  First we look at the integers modulo p, p a prime, (i.e., at GF(p)) so as to introduce basic ideas in a simple setting where examples can be calculated easily. (Indeed, the mathematics here finds application in the construction of Costas Arrays, used in frequency hopping codes in e.g. radar, and in the RSA encryption system.)  Then we consider the arithmetic of GF(2^m), again looking at easily calculated examples. Finally we show how Reed-Solomon codes arise in this setting, again using concrete examples to illustrate the theory.



16 June 2005
Prof. Maja Bystrom
Boston University, USA
"Linking Networks: Coding for Heterogeneous Systems"

Abstract:

As networks increase in size and number of users, optimal or near-optimal resource allocation becomes increasingly important. Furthermore, adaptability in the face of changing channel and source conditions is an interesting challenge. We first review some issues in providing for reliable communications over heterogeneous networks, that is, networks composed of wireline and wireless links. We propose that while serial concatenated codes were designed to provide good overall performance with reasonable system complexity, they may arise naturally in certain cases, such as the interface between two networks. Then, a new method of code rate allocation, based on code performance  bounds, that is flexible in the face of channel variations, is proposed. In particular, we consider the problem of  constrained rate allocation between nonsystematic block codes in a serial concatenated coding system. Given constraints on system parameters, such as a limit on the overall rate, analytic guidelines for the selection of good inner code rates are found by using an upper bound on the average system block error rate.



19 May 2005
Prof. Johannes Huber
University Erlangen-Nuremberg, Germany
"Information Combining: Models, Bounds, and Applications"

Abstract:



11 April 2005
Prof. John B. Anderson
Lund University, Sweden
"From BPSK to MIMO: Thirty Years of Coding and Modulation"

Abstract:
The last 30 years have seen the cost of signal processing drop almost to nothing. This has revolutionized communication and brought us from simple digital modulation to coded modulation to MIMO systems. We will review this history and give a personal view of how one idea led to another and where the technology may go from here.