EPSRC Research Network on Blind Source Separation
and Independent Component Analysis (ICA Research Network)
Partners: Background and Interests
Queen Mary | Aston | Bristol | Birmingham | Cambridge | Cardiff | Edinburgh | Exeter | Glasgow | Goldsmiths | Heriot Watt | Imperial | Liverpool | Manchester | Middlesex | MRC Inst of Hearing Res | Newcastle | Nottingham | Oxford | Paisley | QinetiQ | Reading | St Andrews | Sheffield | Sheffield Hallam | Southampton | Stirling | Sussex | UCL | Warwick
Dr Mark Plumbley (Department of Electronic Engineering)
Dr Mark Plumbley has been working in source separation and independent component analysis (ICA) for 5 years, following a background in neural networks since 1987, including the particular method of non-negative ICA. Dr Plumbley has applied ICA and related models to the analysis of music, including automatic music transcription, and gave the opening keynote talk at the ICA 2004 conference in Granada, Spain. In 2002 Dr Plumbley visited the ICA group at Helsinki University of Technology on a 3-month Leverhulme Trust Study Abroad Fellowship. Recent EPSRC Funding includes: Automatic Polyphonic Music Transcription Using Multiple Cause Models and Independent Component Analysis (GR/R54620/01); Object-based Coding of Musical Audio (GR/S75802/01). Dr Plumbley coordinates the EPSRC Digital Music Research Network (GR/R64810/01).
Prof David Lowe, Prof David Saad (Neural Computing Research Group)
The Neural Computing Research Group is one of the most influential research groups worldwide in neural networks and pattern processing inference methods. It also runs significant research activities in Biomedical Information Engineering and Signal Processing. The Group has championed a principled approach to the theoretical development of neural network structures and architectures and has also developed significant expertise in the engineering applications of its work. Our research linking statistical physics and information theory has led to the development of state of the art error-correcting codes and public key cryptosystems.
Dr Roland Baddeley (Neuroscience)
Dr Roland Baddeley has been working on ICA since his thesis (when it was known as exploratory projection pursuit). As well as developing algorithms, he was the first to apply ICA to the coding of natural images, assessing its biological plausibility by comparison to neurophysiological recordings in cat and monkey, and identifying potential artefacts in this now very popular application. More recently he has been using it to analyse animal camouflage displays, and has been applying it to classic psychological data that had previously been analysed using factor analysis techniques. This area of application is particularly exciting: the ability of ICA to both uniquely identify the rotation, and to find non-orthogonal representations in a principled way, means it is perfect for addressing many of the classic psychometric regularities found in the last 100 years in psychology. In semantics, the classic three dimensional solution of the Osgood semantic differential (1957) appears in part to be an artefact of the imposed orthogonality of factor analysis: two simpler dimensions are derived when ICA is used. Again areas of social psychology research (and particularly fierce controversies) can be simply and directly addressed: ICA provides exactly what standard factor analysis is missing as a useful psychological tool. Such basic problems as the definition of intelligence (often defined based on factor analysis of performance on a bank of intelligence tests), can be greatly simplified; controversies concerning the dimensions of personality are directly addressable. At its most basic, ICA is a method of factor analysis that can remove the arbitrariness of the rotations, and deal with non-orthogonal projections. As such it is a perfect tool for much psychological research. Most if not all the theory already exists for its application to psychological problems. What is required is to popularise it use by both publishing simple applications of the technique to classic factor analysis problems, and writing introductions in a language more familiar to psychologists.
Prof Michael I Friswell (Department of Aerospace Engineering)
Prof. Michael Friswell is the Sir George White Professor of Aerospace Engineering at the University of Bristol. He has a wide experience in problems related to the analysis and identification of vibrating systems, including structural health monitoring. In 1996 he was awarded an EPSRC Advanced Research Fellowship for 5 years, and between 2002 and 2007 he is a Royal Society-Wolfson Research Merit Award holder. Prof. Friswell has published over 100 refereed journal papers, is an Associate Editor of Structural Health Monitoring, of the ASME Journal of Vibration and Acoustics, and of the Journal of Intelligent Material Systems and Structures. He is a member of the EPSRC Peer Review College and among his current EPSRC grants he is a co-investigator on GR/T01136, 'Acoustic Emission for the Localisation of Damaged Regions in Structures as Part of the NDE Process'. Prof. Friswell has used ICA to extract features for health monitoring using vibration measurements, and plans to evaluate ICA and BSS for source location and identification using acoustic emission in NDE.
Dr Ata Kabán (School of Computer Science)
The School of Computer Science of the University of Birmingham is an international leader in the field of Natural Computation (neural networks, evolutionary computation, artificial life) and has recently opened the £2.6 million 'Centre of Excellence for Research in Computational Intelligence and Applications'. It was awarded grade 5 (research quality of international standard) in the last UK Research Assessment Exercise.
Dr Ata Kaban has joined the School of CS of the University of Birmingham in January 2003 as a lecturer. She is a member of the Natural Computation group lead by Prof Xin Yao. Her main research interests concern statistical machine learning and its applications to scientific data mining. She has been working on the development of a general framework of nonlinear latent variable models for the unsupervised & information-preserving visualisation of multivariate data of various types. This model family is closely related to nonlinear ICA models. She has also been working on scalable linear-convex decomposition models of sets of Markovian sequences, with applications to user profiling. Currently, Dr Kaban conducts a small pilot interdisciplinary research project, internally funded by the Paul & Yuanby Ramsay research award scheme, which investigates linear BSS methods for the analysis of stellar population spectra in conditions of measurements uncertainties. This is joint work with Louisa Nolan and Somak Raychaudhury from the School of Physics and Astronomy. Dr Kaban has been a visiting researcher at the Laboratory of Computer and Information Science, Helsinki University of Technology between June-December 2000 and again, with the Data Mining group of the same university in the summer 2003.
Prof Simon Godsill (Department of Engineering)
Prof Simon Godsill leads a group of researchers in Signal Inference and Applications. A particular topic of expertise is audio signal processing, including noise reduction, source separation and music modelling. He also has research interests in Bayesian sequential estimation (`filtering’) using state-of-the-art particle filtering methods, tracking, genomics and Bayesian methods for solving signal inference problems. He currently has research projects funded by the EU, EPSRC, QinetiQ and General Dynamics. He has published extensively in the journal, book and conference literature. His work in the use of Bayesian methods for restoration of degraded sound recordings is world-renowned and he is a director of CEDAR Audio Ltd., a successful commercial spin-off from the audio research at Cambridge. He is an associate editor of IEEE Tr. Signal Processing and the journal Bayesian Analysis, and a member of the IEEE Signal Processing Theory and Methods Committee.
Dr Saeid Sanei (School of Engineering)
Dr Saeid Sanei received his PhD in biomedical signal processing from Imperial College London in 1991. As one of his major directions of research, he has been working in the field of blind source separation and its application to speech, image and biomedical signals for approximately eight years. He has worked in academic institutions and research centres in Iran, Singapore and the United Kingdom since 1991. With effect from 1st September 2004, he will be a Senior Lecturer within the Centre of Digital Signal Processing, Cardiff School of Engineering. He has contributed to the organization of international conferences, such as the IEEE Statistical Signal Processing Workshop 2001, and has served as a reviewer for the foremost IEEE journals.
Prof Mike Davies (School of Engineering and Electronics)
Prof Mike Davies has worked in signal processing and nonlinear modelling since 1989 and held a Royal Society Research Fellowship at UCL and Cambridge from 1993-1998. In 2001 he joined Queen Mary and co-founded the Department’s DSP Group, moving to Edinburgh in 2006. He specialises in nonlinear and non-Gaussian signal processing and has recently made significant contributions frequency domain blind source separation algorithms for audio as well as producing new theoretical and algorithmic results in noisy and over-complete ICA. In 2002 Dr Davies organised the joint IEE/NCAF workshop on “Independent Component Analysis: Generalizations, Algorithms and Applications” hosted at QMUL. He is also an associate editor for IEEE Transactions for Speech and Audio Processing and is currently the guest editor for a forthcoming special issue on Blind Source Separation for the International Journal of Adaptive Control and Signal Processing. EPSRC Funding includes: Advanced Subband Systems for Audio Source Separation (GR/S85900/01).
Dr Richard Everson (Department of Computer Science)
Dr Richard Everson has been working on non-Gaussian Independent Component Analysis and blind source separations problems since 1998, and on related latent variable models of empirical datasets since 1980's. He has introduced new, flexible source models for ICA and has elucidated the manner in which "fixed" source models are able to cope with an unexpectedly wide range of sources. He has developed particle filter methods for tracking non-stationary mixing, and current interests lie in Bayesian methods for non-linear ICA and determination of independent components related to a stimulus signal. He has published several articles on ICA and is co-editor of the book "ICA: Principles and Practice".
Prof Mark Girolami (Department of Computing Science)
Prof Mark Girolami based his PhD thesis (1998) on the development of Independent Component Analysis (ICA) and Source Separation algorithms. He was instrumental in the identification of the underlying causes of the failure of the original 'Infomax' algorithm of Bell & Sejnowski (1995) to isolate artefacts and physiological signals of neurobiological interest when applied to multi-channel electroencephalograph (EEG) data. His development of the 'extended Infomax' algorithm enabled the first successful analysis of multi-channel EEG at the Salk Institute of Biological Studies, and this is now widely adopted as a standard analytical tool employed in EEG analysis. He was funded by the British Library and Information Commission (2000 - 2002) to investigate ICA as a possible alternative to Latent Semantic Analysis (LSA) for text based information retrieval and that work produced results which had impact on the retrieval of images as well. More recently, whilst visiting the ICA group at Helsinki University of Technology, funded by TEKES on a 6 month visiting professor programme, he has studied the problem of overcomplete ICA from a probabilistic perspective and developed variational inference methods to enable the separation of more sources than available observations. His current work at the Bioinformatics Research Centre (BRC) of the University of Glasgow focuses on the study of identifying multiple interacting biological processes from cDNA microarrays and he has recently demonstrated the utility of constrained linear factor ICA based models in identifying biologically valid interacting cellular processes which are phases of the cell cycle.
Prof Michael Casey (Computing Department)
Prof Michael Casey is researching machines that can understand music and other multimedia. His research seeks solutions for automatic organisation of industrial-scale music and video databases. To accomplish this he combines methods of signal processing, pattern recognition and machine learning. Since 1999, he has served as an editor for the MPEG-7 international standard for multimedia content description for which he developed general sound and music indexing solutions for multimedia databases.
Dr Geoffrey Wyatt (Department of Economics)
Geoffrey Wyatt B.A. (Keele), D.Phil. (York), F.S.S. is in the Department of Economics at Heriot-Watt University, where he was Head of Department from 1990 to 1993. He was a visiting researcher at Stanford University in 1994 and at the Research Institute of the Finnish Economy (ETLA) in Helsinki in 1982. From 1974 to 1977 he was seconded to the Department of Economics and Statistics at the OECD in Paris where he produced the OECD's reports on and forecasts of the economies of the UK and Ireland. Major publications: Macroeconomic Modelling in a Causal Framework (En Exempla Books, to appear 2004); Futures and Options: Winners and Losers (Financial Times Business Information Ltd, 1987); The Economics of Invention (Wheatsheaf Press, 1984); Multifactor Productivity Change in Finnish and Swedish Industries (ETLA, 1983). In addition he has produced over 30 shorter academic papers including, most recently: Government Consumption and Industrial Productivity: Scale and Compositional Effects (J.Prod.Analysis, forthcoming) and "Corruption, Productivity and Socialism" (Kyklos, 2004). Recent research awards are: "Population Change in the Regions and Cities of Post-Soviet Russia" (World Bank, 2004); "The Optimal Composition of Government Expenditure: Theory and Evidence" (ESRC, 1999 2002); "Public Finance in Transition Countries: Compliance, Opportunism and Soft Budget Constraints" (EC/ACE 2000 2002); "Applications of A.I. to Economic Modelling" (EC/SPES network, 1992-1995). Current research interests: (i) the representation of economic models as graphs; (ii) data analysis by isotonic regression; (iii) application of independent component analysis to economic data. His main teaching is currently in macroeconomic modelling and analysis; previously he has taught econometrics and mathematical economics.
Dr. Chaoping Zang (Mechanical Engineering Department)
Dr. Chaoping Zang has a BSC in mechanical engineering, an MSc in mathematics and mechanics and a PhD in mechanical engineering. He is currently a Research Fellow at the Mechanical Engineering Department of Imperial College. His main research interests include structural dynamics and testing techniques, rotor dynamics, condition monitoring of rotating machinery, structural health monitoring and damage detection, and applications of artificial neural networks. He has published over 50 papers on structural vibration and machine monitoring problems. Recently he has applied the principal component analysis (PCA) and independent component analysis (ICA) to the structural vibration analysis in order to extract dynamic features for damage detection. The combination of them with ANNs is successfully applied to detect damages in the case of representative structures such as a railway wheel, space antenna, and bookshelf.
Prof Asoke K. Nandi, Dr Judy Zhu (Department of Electrical Engineering & Electronics)
Professor Nandi is the head of the Signal Processing and Communications research group which is composed of some 30 researchers including tenured academics, postdoctoral researchers and doctoral researchers. A significant amount of the research work in the group involves either the development or the applications of independent component analysis.
Prof. Asoke K. Nandi has a longstanding internationally-renowned experience in, among other fields, nonlinear and non-Gaussian signal processing. He has contributed to the area of blind signal processing for over a decade, including fundamental theoretical results as well as the application of blind techniques such as artificial neural networks, independent component analysis and genetic algorithms to machine condition monitoring, communications (channel identification and equalization, automatic modulation recognition, time-delay estimation), acoustics (audio signal separation, underwater sonar, ultrasonic sounding) and biomedical problems (cardiac signal extraction, genomics). He has authored or coauthored over 200 technical publications including two books – entitled Automatic Modulation Recognition of Communication Signals (1996) and Blind Estimation Using Higher-Order Statistics (1999) – and over 100 papers in refereed international journals. In 1983 he was part of the UAI team at CERN (Geneva, Switzerland) that discovered the three fundamental particles known as W+, W- and Z0 providing the evidence for the unification of the electromagnetic and weak forces, which was recognized by the Nobel committee for Physics in 1984. He then held a number of positions at the University of Oxford, Imperial College (London) and the University of Strathclyde (Glasgow). Since 1999, he is the David Jardine Professor of Electrical Engineering in the Department of Electrical Engineering and Electronics, The University of Liverpool. He heads the Signal Processing and Communications Research Group, which includes a number of tenured academics and a large group of doctoral and post-doctoral researchers with diverse interests in the areas of nonlinear systems, non-Gaussian signal processing and communications. He was awarded the Mountbatten Premium, Division Award of the Electronics and Communications Division, of the Institution of Electrical Engineers of the U.K. in 1998 and the Water Arbitration Prize of the Institution of Mechanical Engineers of the U.K. in 1999.
Prof Jonathon Chambers (Department of Electronic and Electrical Engineering)
Professor Jonathon Chambers has been working in the area of blind signal processing for a number of years and has contributed to the development of the field. He is an active collaborator with Professor A. Cichocki at the Brain Science Institute (BSI) within RIKEN, Japan, which was established in 1997, well-known as an international research centre of excellence in brain science. In October 2003, he delivered an invited plenary talk entitled “Open Issues in Constrained Blind Source Separation” at the IEEE New Directions in Signal Processing in the 21st Century Workshop, Chateau Lake Louis, Alberta, Canada. He is currently the PI on two active EPSRC projects related to this proposal: Blind Signal Processing for Multi-channel Speech Enhancement, GR/R69228/02, and Room Acoustic Parameters From Music, GR/S77547/01. He is currently an Associate Editor for IEEE Signal Processing Letters, and the EURASIP Journal on Wireless Communications and Networks, and is about to serve for a second term as an Associate Editor for IEEE Trans. Signal Processing.
Dr Magnus Rattray and Dr Neil Lawrence (School of Computer Science)
Dr Magnus Rattray works on machine learning and probabilistic modeling methods with recent applications in bioinformatics. From 1996-1998 he worked as an EPSRC funded research fellow at the Neural Computing Research Group, Aston University, working on the theoretical analysis of on-line learning in neural networks. In 1998 he joined the Computer Science Department, Manchester University, as a Lecturer and later Senior Lecturer. During this time he developed a novel theoretical framework for the analysis of on-line ICA, with this work being published in NIPS, ICANN, Neural Computation and JMLR. Current interests focus on the application of statistical learning algorithms to large-scale data sets produced by recent genomics and functional genomics technologies. Dr Rattray has been a PI on grants funded by the EPSRC (Analysis of Natural Gradient Descent for Statistical Models - GR/M48123/01), BBSRC and industrial sources.
Dr Neil Lawrence is a Senior Research Fellow in the School of Computer Science. Until 2006 he was a Senior Lecturer in the Department of Computer Science at the University of Sheffield. Previous to this appointment he spent a year working on a post-doctoral contract with Microsoft Research (MSR) in Cambridge, U.K.. This position followed on from his PhD that was awarded by St John's College, Cambridge. Dr Lawrence's research has focused on analysis of genomic data with probabilistic models, where he collaborates extensively with the University’s bioscience departments, and probabilistic models for speech recognition, where he collaborated with the Department's Speech and Hearing group. Dr Lawrence currently supervises four PhD students, mostly working in speech.
Dr Hujun Yin (School of Electrical and Electronic Engineering)
Dr Hujun Yin works at UMIST Vision and Information Processing group and his research interests include theories and applications of neural networks, self-organising learning systems in particular, independent component analysis, data mining and visualisation, image processing, compression, and bioinformatics. His main contributions are in convergence theory of self-organising systems, mixture models, and nonlinear data projections. Recent EPSRC grants include: Bayesian and Fault Tolerant SOMs; Nonlinear ICA for Image Denoising. Currently he has a BBSRC grant on Applying Machine Learning for Protein Peptide Identification from Mass Spectra. Dr Yin has been involved in an early EPSRC network on Emergent Computing in 2000-2001. He has organised a number of international conferences: WSOM’01, IDEAL’02 and IDEAL’04, and has been an invited guest editor for Neural Networks in 2002.
Dr Roman Belavkin has become interested in ICA while teaching an AI course on MSc in Business Information Systems programme at Middlesex University. Although his main interests are in cognitive modelling of decision making and learning, he has background in physics and he implemented several algorithms for multivariate data analysis with particular interest to business problems. During 2003--2004, he supervised several MSc projects, in which ICA was used to analyse various data sets, such as hotel occupancy rates and currency exchange rates. Belavkin received his MSc in physics from Moscow State Univeristy (1994) and PhD in computer science from the Univeristy of Nottingham (2003).
Dr Deb Hall, Dr Silvia Cirstea
The MRC Institute of Hearing Research has a core research group who use neuroimaging techniques (fMRI and EEG) to investigate the functional organisation of the human auditory brain. Our work has defined a broad taxonomy of auditory cortical subdivisions that are sensitive to the spectro-temporal properties of sound. Advanced data acquisition techniques and analysis methods are essential to support sophisticated experimental investigation of human brain function and so we strive to employ contemporary analysis methods (independent components analysis, structural equation modelling etc) in addition to general linear modelling.
Dr Deb Hall leads the neuroimaging research program in human auditory fMRI at IHR. Dr. Silvia Cirstea, has worked in mathematical modelling with application to image processing, acoustics and radio wave propagation. She specialises in inverse problems in imaging and has reported regularization methods for depth extraction and object space reconstruction from 3D-integral images.
Dr. Wai Lok Woo (School of Electrical, Electronic and Computer Engineering)
Dr. Wai Lok Woo is a full-time academic staff in the School of Electrical, Electronic and Computer Engineering at the University of Newcastle. He has a portfolio of relevant research supported by a variety of funding agencies. He was the recipient of the IEE Student Prize in 1998 and awarded the British Scholarship in the same year to continue his research work at Newcastle University. Prior to joining the School, he works on Blind Source Separation techniques which were supported by QinetiQ Limited on signal processing-based applications. The work were diversified to investigate and develop new tools for analysing nonlinear mixtures. One new finding is the proof of a previously missing link that substantiates the use of multiple hidden layer neural network for Nonlinear Independent Component Analysis. Since joining the University, he has continued to engage in nonlinear blind signal processing techniques for signal restoration, channel identification and equalisation, geophysical sciences, biomedical and speech processing. Dr. Woo was the local organiser of the 4th International Symposium on Communication Systems, Networks and Digital Signal Processing which was held in UK. He is also a member of IEE, IEEE and SPIE.
Dr Malcolm Woolfson (School of Electrical and Electronic Engineering)
Dr Malcolm Woolfson is a Lecturer in Signal Processing at the University of Nottingham. He has been involved in the application of signal processing methods to several areas of engineering, including biomedical engineering, power systems, sensorless speed determination of motors and electromagnetic compatibility. One of his research areas is the development of Blind Source Separation methods for signal mixtures where the sources are sparse. The approach used here is to look at local parts of phase space where one source dominates and to use this information to extract each source. One application where this could be used is the extraction of the fetal ECG from maternal abdominal and thoracic data
Prof Stephen Roberts (Department of Engineering Science)
Stephen Roberts is head of the Pattern Analysis & Machine Learning Research Group, University of Oxford. He has some 15 years experience in the fields of signal processing, pattern analysis and machine learning. Stephen gives undergraduate lectures in mathematics, probability theory, pattern analysis and signal & image processing as well as lectures in machine learning to MSc students. Stephen's main interests lie in the application and development of mathematical methods to data analysis and data-driven machine learning, in particular Bayesian learning and inference. He has supervised 20 PhD students and authored / co-authored over 120 refereed publications including two books (Independent Component Analysis: CUP, 2001; Applications of Probabilistic Modelling in Medical Bioinformatics, Springer, 2003). He has been awarded over £2 million in grants (£1.5m in the last three years). He has been awarded IEE medals for two years running for papers on applied inference and has published widely in the area of ICA. With co-authors he has some 20 publications, ranging from mixture of ICA analysers to Markov chain ICA models. His recent work has concentrated on the use of hybrid ICA-wavelet models and the use of ICA to perform unsupervised data segmentation. The Department of Engineering Science has a 5* grading from the last RAE and an environment capable of excellent research. There is considerable expertise in the department in the areas of data analysis, machine learning and signal processing. Dr Roberts has links with machine learning and data analysis researchers both nationally and internationally.
Dr Christian F. Beckmann (Department of Clinical Neurology)
Dr. Christian Beckmann has been working on the application of Independent Component Analysis to functional imaging data since 1999, initially as part of his DPhil at the Medical Vision Lab, Department of Engineering, University of Oxford and since 2002 as a Research Officer at the Department of Clinical Neurology, University of Oxford. His research has lead to ICA software which now is being used within the international FMRI community as part of the FMRIB Software Library, a collection of tools for the analysis of functional and structural MRI data. His current research focuses on novel applications of ICA to neuroimaging data, regularisation approaches to ICA as well as new algorithms for multi-way generalisations of ICA for the analysis of higher-dimensional data sets.
Dr Irene M Moroz, Max Little (Oxford Centre for Industrial and Applied Mathematics)
Applied Dynamical Systems is an interdisciplinary mathematical research group based at the Oxford Centre for Industrial and Applied Maths, Oxford University. The group collaborates with the Engineering Science and Atmospheric Physics departments of the university. The time series that the group study originate in several types of systems for which measurements are tractable or abundant, for example, voice recordings, electrical circuits and atmospheric data. The research covers signal analysis, modelling and synthesis using applied general dynamical systems theory, general basis transforms, machine learning, information theoretic and complex systems methods. Current projects include voice morphing, low-dimensional modelling of the Martian atmosphere, nonlinear signal processing and studying nonstationarity in chaotic electronic circuits. The group has used PCA and ICA to find sparse bases for voice representation and to construct low-dimensional atmospheric models. Specifically they intend to pursue the use of these methods for more detailed work on atmospheric modelling.
Prof Colin Fyfe, Prof Douglas Campbell (Applied Computational Intelligence Research Unit)
Prof Fyfe has supervised 15 completed PhDs since 1998 and one theme of these is the development of ICA algorithms from an artificial neural network perspective. This is apparent in the contents of a new book, "Hebbian learning and negative feedback neural networks" (Springer, 2004) which is drawn from the work of these PhDs. The Applied Computational Intelligence Research Unit at the University of Paisley is composed of about 12 lecturers many of whom are involved in ICA research: for example, Prof Douglas Campbell has used ICA techniques to improve hearing aids.
Prof John McWhirter, Mr Paul Baxter (Advanced Signal and Information Processing)
QinetiQ's Advanced Signal and Information Processing group has been developing BSS algorithms since the early 1990's. The group has applied BSS techniques to scenarios ranging from radar and comms to biomedical applications, and current research is focused on the BSS of convolutive mixtures and on non-stationary (tracking) BSS. The group is headed by John McWhirter FRS.
Dr Slawomir Nasuto, Dr Nicoletta Nicolaou (Department of Cybernetics)
Dr Slawomir J. Nasuto is a lecturer at the Department of Cybernetics, University of Reading. He has been working on biosignal analysis for 3 years with a previous experience in signal processing using Cybernetic Intelligence methods and Computational Neuroscience since 1995. Dr Nasuto is interested in application of such techniques including Independent Component Analysis (ICA) in biomedical image analysis, electroencephalography and electromyography with emphasis on Brain Computer Interfaces and potential medical diagnostic applications.
Dr Nicoletta Nicolaou has a background in Cybernetics and Control Engineering, and is currently working on the application of ICA in the analysis of EEG signals. In particular she is investigating the application of ICA as a technique for pre-processing, automatic artefact removal and feature extraction from EEG signals. She has been working with ICA and various extensions, mainly temporal, of the basic technique since the start of her PhD.
Dr Ines Jentzsch, Dr. Peter Földiák (School of Psychology)
Dr. Ines Jentzsch is a trained Biophysicist working in the field of Experimental Psychology and Cognitive Neuroscience, and supplies expertise in the source localisation and dipole analysis of event-related potentials (ERPs), including the application of Independent Component Analysis (ICA). Dr Jentzsch has been trained in applying ICA to ERP analysis in November 1998 by S. Makeig and T.P. Jung at the Salk Institute, San Diego, USA and has been using the method since for her research in the fields of human information processing, cognitive control and movement planning. Dr. Jentzsch was invited to the workshop “Analyzing and Modelling Event-Related Brain Potentials - Cognitive and Neural Approaches”, held at the University of Potsdam in Fall 2001. She also received a Postdoctoral Research Fellowship from ESRC at Glasgow University from 2002 to 2003 and now holds a permanent academic position at the School of Psychology, University of St Andrews.
Dr. Peter Földiák has a background in electrical engineering, computer science, and physiology. He studies biological mechanisms and neural network models of sensory information processing with special interest in vision and pattern recognition. He is also interested in perceptual adaptation, semantic knowledge representation and artificial intelligence.
Dr Jim V Stone (Psychology Department)
Dr Jim Stone is interested in source separation problems as a means of solving problems in diverse research areas including sound separation, fMRI analysis, preprocessing for cerebellar models, and especially unsupervised learning of perceptual invariants. He has devised novel forms of BSS methods, of which spatiotemporal ICA has been applied to fMRI data, and his BSS algorithm based on temporal predictability has been applied to unsupervised learning of visual stereopsis, and the deconvolution and separation of sound signals. He has written two review articles on ICA and a book, which is due for publication in the autumn of 2004.
Prof Keith Worden (Department of Mechanical Engineering)
The dynamics group is active in signal processing as applied in structural dynamics. Prof Warden is head of the Dynamics group: his main interests are in nonlinear systems and in machine learning approaches to structural health monitoring. He is interested in ICA as a possible way to unravel signals from ultrasonic testing.
Dr R Saatchi (School of Engineering)
Dr R Saatchi has been active in the field of signal processing since 1988. He did his PhD in biomedical signal processing by analysing electroencephalogram (EEG) to monitor specific brain disorders. One of his papers in this area was awarded the IEE Premium Prize in 1995 and the EUREL prize in 1997. His research activities have included devising a multi-resolution analysis based filtering technique for recovering evoked potentials buried in the electroencephalogram waveforms and using a number of signal source separation approaches to unmixed and characterise EEG signal components. He has supervised two successfully completed PhDs in these areas.
Dr Christopher J. James (Institute of Sound and Vibration Research - ISVR)
Dr Christopher James has been working in the field of biomedical signal and pattern processing since 1993. He has held research fellowships at the Montreal Neurological Institute and the Neural Computing Research Group at Aston University in Birmingham. He has been recently appointed lecturer in biomedical signal processing at the ISVR at the University of Southampton. He is on the relative biomedical engineering executive committees of both the IEEE and the IEE and is the biomedical engineering series editor for Artech House Publishers. He established and co-chaired a special session on ICA applied to brain signals at the IEEE EMBS Annual Conference in Turkey 2001. He has authored invited papers on ICA/BSS applied to brain signals and a review paper (in press) in ICA/BSS applied to physiological signals. He is actively involved in research to establish and improve ICA/BSS techniques in biosignal analysis – especially in the field of epilepsy. EPSRC funding includes: Advancing the state-of-the art in early seizure onset analysis in the EEG (GR/S13132/01).
Dr Manfred Opper (Image, Speech and Intelligent Systems - ISIS)
The Image, Speech and Intelligent Systems (ISIS) group activities are centred in fundamental theory and algorithm development associated with adaptive data modelling, machine learning, control theory and signal processing. This research is developed through the verification and validation in the real-world problem domains of vision and image processing, speech, guidance and control, autonomous systems command+control, automotive, aerospace, biomedical fields, etc. The group is part of the School of Electronics and Computer Science that received top ratings in both Electronics and Computer Science at the last RAE.
Manfred Opper obtained a PhD in physics in 1987 in Giessen, Germany. After that he began to work on the application of statistical physics to the theory of neural networks. Following postdoctoral visits at the Ecole Normale Superieure in Paris (1989) and at the University of California at Santa Cruz (1990) he received the habilitation degree in theoretical physics in 1991. In 1992 he was awarded the Physics Prize of the German Physical Society for work on the dynamics of learning in neural networks. He was awarded a 3 year Heisenberg fellowship in 1994 enabling him to work at the machine learning group of UC Santa Cruz and at the Department of Complex systems of the Weizmann Institute in Israel. He joined the Neural Computing Research Group at Aston University as a Reader in 1997. In 2004 he moved to Southampton University where he is a reader at the ISIS research group. He has over 110 publications, most of them in the application of statistical mechanics methods to problems of machine learning and other complex systems. His research interests are in the cross connections between statistical physics, computational learning theory, information theory and mathematical statistics. He is currently working on the theory and applications of Gaussian processes and on the development of efficient inference techniques for complex probabilistic models. He was the principle investigator of EPSRC grant GR/M81601 (overall assessment: outstanding) which analysed kernel methods in machine learning using approaches from statistical physics and was the Co-Investigator of EPSRC grant GR/R61857/01.
Dr. Amir Hussain (Computational Intelligence Research Group)
Amir Hussain, B.Eng. (1st Class Honours), Ph.D (Strathclyde), is a Senior Lecturer in Computing Science and a lead member of the Computational Intelligence Research Group at the University of Stirling. His research interests are inter-disciplinary and include: neurobiologically motivated modeling and control of complex and cognitive systems, machine learning and applications to data mining and audio signal processing, including noise reduction and source separation. His research activities have been supported by, amongst others, the EPSRC, EU, Royal Society and industry and these have led to over 100 (co)authored publications to-date. He was PI and co-ordinator of a recent EPSRC Novel Computation Research Network on Learning Control Paradigms for Complex Systems and was Stirling PI of a recent EU project on intelligent decision support systems. He is Chair of the IEEE UKRI Industry Applications Society Chapter and is Associate Editor of the Control & Intelligent Systems Journal, and the International Journal of Robotics & Automation. He is invited Guest Editor of the Neurocomputing (Elsevier) Special Issues on Brain Inspired Cognitive Systems and is founding co-Chair of the IEEE International Conference on Engineering of Intelligent Systems. He is currently an invited expert member of the European (ESF) networked research project on cross-modal analysis of verbal and non-verbal communication and is leading a new collaborative project on audio-visual speech processing for noise reduction and source separation.
Daniel Osorio, John Anderson (School of Life Sciences)
Animal behaviour involves co-ordinated activation of large numbers of muscles under neural control. Because of the complexity of much behaviour is difficult to analyse the organizational principles. One might hypothesise for instance that there are a limited number of neural ‘centres’ which can be more or less independently activated to produce a flexible range of actions. Such a model is particularly relevant to signalling behaviours. For instance Charles Darwin (1872, The Expression of the Emotions in Man and Animals. London, John Murray) noted that human emotional can expression is based on mixtures of a few basic types, such as fear, disgust and happiness. The study of behavioural co-ordination and biological signalling will benefit greatly from methods for identifying the separate causes of sources of a signal, such as ICA, because this may allow us to understand how complex outputs are generated. We are using ICA for investigating coloration patterns of the common cuttlefish, Sepia officinalis. This animal flexibly combines a large number of separate motifs or elements for camouflage and signalling. These patterns can be altered in fractions of a second. Published work has demonstrated that the method is effective for a small image datasets, and we are currently collecting further, and much more detailed data on coloration in a range of behavioural contexts. More generally ICA is considerable appeal because compared to techniques such as PCA independent components identified are more intuitively comprehensible and meaningful. Thus it should find a much wide range of uses in biology and neuroscience. ICA is so far poorly known and an EPSRC funded network will be of great value in development of methods and applications.
Prof Bernard Buxton (Department of Computer Science)
ICA methods are of interest to the group and is relevant to some of our pattern recognition and computer vision research. We have interest in their application to bioinformatics problems and, to a greater extent, to a variety of computer vision problems, including: (i) human face and expression modelling and analysis, in particular in conjunction with our recently developed integrated shape and pose modelling methods (ISPM); (ii) texture modelling, eg for skin (and, if possible, hair) for the appearance of the human face; (iii) the tracking of human figures from monocular camera data, in particular re: (a) modelling the shape of an articulated object such as the human figure, and (b) modelling the temporal correlations in the motion of such figures, with (c) application to computer enhanced dance performance and entertainment (in collaboration with the University of Brighton and the University of Surrey, Roehampton College), (iv) the modelling and analysis of human 3D whole body scan data. Prof Buxton trained as a theoretical physicist, and also has a personal interest in the theory of ICA methods.
Dr Christophe Ladroue (Department of Statistics)
Dr Christophe Ladroue's initial training is in mathematics. He obtained his PhD from St George's Hospital Medical School, university of London, for his work on pattern recognition techniques for brain tumour NMR spectroscopy data, which was funded by the European Community (INTERPRET, IST-1999-10310). An important part of his thesis was dedicated to the use of ICA for the automatic decomposition of complex MR spectra and results from this study were published as a full paper in Magnetic Resonance in Medicine (50(4):697-703, 2003). He worked for Prof. John Griffiths' MR research group at SGHMS and was a visiting research fellow at the university of Sussex in the informatics department. One of his interests has been to refine the ICA method he has used so far and tailor it for MRS data.
This page updated 18-Oct-2007