[I5] Synthetic Noise
The capability to generate spatially and temporally correlated non-stationary noise is becoming an increasingly important requirement for the assessment of advanced signal processing algorithms in the laboratory, for examination of usefulness of virtual sensor and for factory acceptance tests (FATS) of equipment. In the past, common methods of performing laboratory simulations utilised either measured data and/or spatially uncorrelated Gaussian noise. However, each of these methods have advantages and disadvantages for their use in signal processing schemes. For example, whilst white Gaussian, spatially uncorrelated noise is simple to generate, it suffers from a lack of realism. Measured noise includes realism, but it has other disadvantages such as; the potential for unknown events to occur; the inability to perform many numbers of tests and it is usually expensive to obtain. Therefore, there is a need for an intermediate stage of synthetic noise generation which retains the advantages of uncorrelated noise and real data whilst removing some of the disadvantages. This requirement is becoming more significant since signal processing schemes are becoming increasingly sophisticated and the emphasis is moving away from collecting real data due to cost/risk; sometimes real data is impossible to collect. And synthetic noise is even more important. This proposal is aimed at the requirement to produce synthetic non-stationary spatial/temporal noise.
Project Supervisor
Timothy Clarke, Defence Science and Research Laboratory (Dstl). Tim Clarke has a wealth of experience in acoustics, particularly in the underwater domain, having worked in academia, private industry and government research laboratories. One area of Tim’s expertise is the link and exploitation between the environment and signal processing for acoustic systems.
Project Summary
Project Type: Accepted Status: Open Call[O12] Real Time Model Adaptation for Non-Stationary Systems
This project investigates new approaches of modelling for the non-stationary data sets, which are commonly generated in many systems including Radar, Sonar, communications, instrumentation, seismic exploration, speech processing and recognition, etc. Signal processing functions usually perform based on a pre-set model, or the system structure is fixed. Although this provides a simple solution, it is highly inefficient especially for non-stationary systems that are common in practice. This makes it highly desirable to adapt the model so that it can capture the true underlying dynamics and predicts accurately the output for unseen data.
Against this background, however, there is a lack of generic tools/ methodologies to deal with the problem as on how to perform the model structural changes as demanded by the processes. Some nonlinear model structure identification algorithms are too slow for real-time applications. Most current real-time algorithms, on the other hand, are ad hoc rather than principled approaches. Hence the resultant model quality is not optimal from a statistical point of view.
In this programme, we propose to develop a hybrid, flexible yet principled approach for optimum on-line adaptive modelling by means of minimal model structure determination and simultaneous parameter estimation. The aim of the proposal is to introduce a new technique for the adaptive modelling of complex nonlinear dynamical systems in real time and noisy environments.
Project Supervisor
Dr Gong obtained BEng and MEng at University of Electronic Science & Technology of China in 1992 and 1995 respectively, and PhD from National University of Singapore in 2002. From 2002 to 2003, he worked firstly as an engineer, and later a senior engineer, at Institute of Inforcom in Singapore. In 2003, he joined Queen’s University of Belfast as a Research fellow, and later took an engineer post at the Institute of Electronics, Communications and Information Technology (ECIT), QUB in 2005. Since 2006, Dr Gong has been with University of Reading as lecturer. His research interests include communications, signal processing, adaptive filtering etc.
Project Summary
Project Type: Accepted Status: Open Call[I3] Extracting tones which vary in frequency from non Gaussian noise
The project will the address the UDRC challenge: "To distinguish between man-made echo-sounding pulses and those made by marine mammals (cetaceans)"
Ever since the early days of passive sonar (receive only) there has been a need to determine whether sounds in the ocean emanate from the native marine life or from mechanical sources which may turn out to be contacts of interest. Furthermore as modern Navies aspire to be ever more responsible environmental custodians of the underwater environment, it is necessary to have a situational awareness of local biological activity before any active sonar (transmit and receive) transmission. The idea is to develop frequency tracking schemes tailored to the underwater environment with specific regard to the tracking of cetacean tonal vocalisations. The approach will be block based and take advantage of the fractional Fourier transform to feed the particle filter with estimates of both centre frequency and chirp-rate. The Matlab based algorithms will be benchmarked against real datasets and the current best of breed.
Project Supervisor
Dstl Naval Systems, Portsdown west, Portsmouth, PO17 6AD. Tel: 023 9253 2668,
Project Summary
Project Type: Accepted Status: Open Call[C6] Flexible Array Signal Processing
The aim of this project is to carry out comprehensive investigation of sensor-arrays where the position of the individual sensors is described by a time varying function, that is the geometry changes during the observation interval. Thus the proposed work is concerned with the development of novel suitable superresolution flexible array signal processing techniques for accurate detection, localisation and tracking of narrowband and/or wideband targets in complex environments. This objective is motivated by the military needs for the flexible sensor arrays (e.g. radar, sonar) that have not only high resolution location estimation capabilities but also robustness against the unwanted effects of jammers and noise.
In general, two different types of time-varying sensor-arrays can be distinguished. The first type is "rigid" array of sensors mounted on mobile platforms. In this type, the geometry of the array is preserved at least during the observation interval - if not throughout the movement of the platform. The second type is "flexible" array of sensors where the array geometry changes during the observation interval of the incoming signals from the targets. The geometry of the problem in the first type is equivalent to the case of a fixed array and a moving source since only the relative motion of the source and the array is of importance in this case. While the problem formulation of this proposal is quite general, particular emphasis will be placed on the second type of array systems namely where the shape of the array changes significantly during the observation interval.
Within the type of flexible sensor arrays, two categories of movement can be identified. Firstly, the sensors of the array may change their positions in an arbitrary but known way. A good example of this movement is sensors attached to Unmanned Aerial Vehicles (UAVs) where each sensor element has its own propulsion system and they group together to form a large aperture array of moving sensors. In the
second category, the geometry of the flexible array may change in an unpredictable and immeasurable way. For example, due to imperfections in vehicle control, it is common for UAV arrays systems to undergo significant yaw and pitch oscillations. Another example is large towed hydrophone arrays in real situations where sensors may be displaced from their nominal positions due to various kinds of forces on the array. In these examples, time varying uncertainties are introduced in the sensor locations that may degrade the performance of the array system and can only be described stochastically. In both types of movement, the motion of flexible array elements causes the array manifold (i.e. the locus of all sensor array responses) to be a function of time.
Project Supervisor
Project Summary
Project Type: Accepted Status: Core Research[C3] Widely Linear Adaptive Processing of Noncircular Complex Signals
This research will provide a novel theoretical framework and enhanced practical solutions for adaptive processing (supervised and blind) of noncircular complex signals. Standard, widely used, solutions inherently assume second order circularity of signal distributions, and are therefore inadequate when the signals are observed through nonlinear sensors or as mixtures of sources, when the noise model is not doubly white, and when high resolution and enhanced separability are paramount. This will be achieved based on recent fundamental developments in the statistics of complex variables, called augmented complex statistics. The fundamental novelty in this work is the design of statistical signal processing techniques for the detection, estimation, and quanti cation of the degree of noncircularity of real world signals. This information will be used in conjunction with widely linear adaptive signal processing techniques, to enable optimal processing of complex signals with both circular and noncircular distributions.
The outcomes of this research will have an immediate impact in a variety of applications of signal detection, estimation, and adaptive signal processing, including those in interference supression, direction of arrival estimation, and blind extraction of sources of interest in real world scenarios.
Project Supervisor
Dr. Mandic received the Ph.D. degree in nonlinear adaptive signal processing in 1999 from Imperial College, London, London, U.K. He is now a Reader with the Department of Electrical and Electronic Engineering, Imperial College London, London, U.K. He has previously taught at the Universities of East Anglia, Norwich, Norfolk, U.K., and Banja Luka, Bosnia Herzegovina. He has written over 150 publications on a variety of aspects of signal processing and a research monograph on recurrent neural networks. He has been a Guest Professor at the Catholic University Leuven, Leuven, Belgium and Tokyo University of Agriculture and Technology (TUAT), and Frontier Researcher at the Brain Science Institute RIKEN, Tokyo, Japan. Dr. Mandic has been a Member of the IEEE Signal Processing Society Technical Committee on Machine Learning for Signal Processing, Associate Editor for IEEE Transactions on Circuits and Systems II, and Associate Editor for International Journal of Mathematical Modeling and Algorithms. He has won awards for his papers and for the products coming from his collaboration with industry.
