Poster & Presentation Abstracts for the Research Open Day 2008

Research Groups:

1: Antenna and Electromagnetics

2: Centre for Digital Music (C4DM)

3: Multimedia & Vision

4: Networks

1: Antenna and Electromagnetics Research Group

Validation of Quasi Optical Design Tool for Quantitative Analysis of Beam Aberration Caused by Planar Signal Conditioning Components

By: Anoma Yamsiri

One of the most important concerns for optical system used in radiometer is beam aberration that degrades accuracy of the system. Methods of beam aberration analysis at a focusing reflector are well known but the aberration occurs at a planar signal conditioning component such as frequency selective structure or wire grid polarizer has received less attention. Therefore, quasi optical (QO) circuit is designed for analysis of beam aberration caused by such devices. A validation of QO design tool is the subject addressed in the presentation.

Unit-Cell Characterisation of Metamaterial Based Leaky-Wave Antenna

By: Atiqur Rahman

Transmission line (TL) approach of left-handed (LH) metamaterial, using non-resonant and therefore, low loss and broad bandwidth structures, was proposed almost simultaneously by a number of groups. Several microwave concepts and applications have been developed based on this approach. One of the most promising applications of composite right/left handed transmission line (CRLH-TL) structure is the leaky-wave antenna (LWA) which has a superior frequency scanning ability than its conventional counterpart. Although periodic type leaky-wave antenna can produce backward radiation, its broadside radiation is restricted by the stop band between the forward and backward region. While the conventional periodic-type leaky-wave antenna lacks good broadside radiation, CRLH-TL leaky-wave antennas can have improved broadside gain provided that the balanced condition is satisfied. Another attempt to implement CRLH-TL leaky-wave antenna was made with CPS technology, however, only forward beam direction is shown and also balanced condition is not satisfied here. Although several designs of CRLH-TL based LWA are available in the literature, considerations for the unit-cell design have not reported. Here, we have investigated the relationships between the different designs of the unit-cell and the performance of the CRLH TL based leaky-wave antenna in terms of bandwidth, scan-range and antenna gain by calculating the radiation resistance and analysing the dispersion characteristics of the designed antenna. It has been shown that the leakage rate of the CRLH-TL LWA can be altered by using different sizes of the unit-cell.

Assessment of Magnetic Materials for Use in Quasi-Optical Non-Reciprocal Devices Operating at Frequencies above 90 GHz

By: Bin Yang

High-performance quasi-optical non-reciprocal devices (isolators, circulators) are required for use in many current and future measurement systems operating at millimetre-wave/submillimetre-wave frequencies. In such a device, free-space quasi-collimated linearly-polarised signal-beams pass normally through a plane plate of a magnetic material that is uniformly magnetised perpendicularly to its surfaces and is of a thickness that will give a 45-degree Faraday rotation of a linearly polarised signal beam in a single pass.

The material of the plate must therefore have the static magnetic characteristics of an excellent permanent-magnet material to retain the uniform magnetisation at near-saturation value in the presence of the large demagnetising field.  The best commercially available non-metallic permanent-magnet materials (the hexaferrites, metal-oxide oriented ceramics) have the necessary axial anisotropy and near-rectangular demagnetisation curves; few of them, however, prove to have a large enough ratio of coercivity to remanent-magnetisation to countervail such large demagnetising fields.

A candidate material must also have the necessary gyrotropic magneto-optical properties at millimetre-wave/submillimetre-wave frequencies. However, most of the commercially available hexaferrites attenuate millimetre-wave/submillimetre-wave beams too strongly.

It is therefore necessary to screen samples of permanent-magnet materials from many suppliers in order to find good candidate magnetic materials for use in millimetre-wave/submillimetre-wave non-reciprocal devices. This thesis establishes the selection criteria for candidate gyrotropic material first, and then designs the novel quasi-optical broadband measurement bench and presents the fast and accurate measurement methods and procedures to measure candidate hexaferrites.

Dynamics in Plasma-Filled Diode with Loads in the Presence of a Magnetic Field

By: Daohui Li

The operation of microwave generators is based on an interaction process between electrons and an electromagnetic field, the kinetic or potential energy of the electrons being gradually converted into the radio frequency energy of the field. The process is highly nonlinear, which is capable of exhibiting very complicated dynamical behaviour. Since the mechanism of the nonlinear dynamics, especially chaos, is difficult to be identified clearly in real microwave generators. A plasma-filled planar diode, which has an injected electron beam and a stationary cloud of positive ions set up between two parallel planar electrodes, can be taken as a convenient model for some microwave oscillators or amplifiers.

Our study has considered the effects of some loads connecting between the two electrodes in presence of a magnetic field in the plasma-filled diode. The nonlinear dynamics of the system can be investigated under suitable combinations of some key control parameters, i.e. the ratio of ion to electron density at the entrance electrode, the separation of the electrodes, the magnetic field and the load. Four distinct dynamical regimes: steady, oscillatory, chaotic and unstable, have been achieved via both mathematical analysis and particle-in-cell simulation.

Millimetre Wave Imaging for Concealed Weapon Detection

By: Lianhong Zhang

Concealed weapon detection (CWD) has been a hot topic as the concern about public safety increases. Millimetre wave imaging is considered as a promising approach due to the relatively high penetration ability and high resolution it can offer. The millimetre wave imaging system is analogous to an optical imaging system. The pupil plane scanning system is mainly discussed in this presentation. Compared to the image plane scanning system, it has the advantages of high reliability and simple structure design. Such a system can be passive or active and operate at a single frequency or multiple frequencies (wide band). An active, coherent, single frequency experimental millimetre wave imaging system is designed based on the theory of synthetic aperture and holographic processing. This system allows validating the design without actually building the costly system. Imaging array spacing is an important system parameter and it is investigated. Further investigations such as imaging system resolution, optimized system transmission power, working range etc. will be followed up.

Miniaturisation of Printed Disc UWB Monopoles

By: Lu Guo

The Ultra-Wideband (UWB) radio technology has been attracting more and more attention recently since the approval by the Federal Communication Commission (FCC) of a 3.1-10.6 GHz frequency band allocation. As a critical part of the whole system, UWB antennas have been receiving considerable interests from both the academia and industries. During the past few years, many types of UWB antennas have been proposed and studied and some progresses have been made towards miniaturisation of planar UWB antennas. In this paper, the miniaturization of a printed circular disc monopole antenna is investigated. It is shown that by halving the original antenna and adjusting the coplanar ground plane width, an even wider impedance bandwidth at the lower end can be obtained. Furthermore, a clear physical explanation is provided to illustrate the mechanism of this type of half circular disc antenna.

Role of Surface Waves in On-Body Bluetooth Link

By: Masood Ur Rehman

The human body can be used as a communication channel for different devices located in the close proximity of the body as well as body-worn devices. A Bluetooth enabled body-worn mobile handset and the Bluetooth headset form a wireless communication link. On-body wireless communication link involves two mediums for the transfer of signals, i.e. the human body and the air. It has been noticed that when there are no line of sight and reflecting objects, electromagnetic waves are transferred between two body-mounted antennas mostly via the interface between the body and the air in the form of surface waves. This study presents an investigation into the on-body Bluetooth transmission mechanism using a simplified homogenous human body model. The single layer homogeneous human body model was constructed with an averaged relative permittivity of 28.16 and conductivity of 1.14 S/m at 2440MHz, which is compounded considering 10% skin, 30% fat, 40% muscle and 20% bone to make it near to actual. Antenna models used in this study are based on the actual antennas available in commercial products.

The study consists of the handset PIFA (Planar Inverted-F Antenna) and meandered line monopole headset antennas modeling, simulations and measurements of the Bluetooth link budget between the two antennas at 2440MHz in free space and on-body in two orientations and a thorough analysis of the surface waves by observing electromagnetic wave patterns and examining electric field distributions.  Effects of the separation between the body and the handset antenna and a horizontal metal barrier are also studied to further check the role of surface waves.

Analysis of Mutual Coupling in Broadband Arrays

By: Sheng Wang

The electromagnetic performance of an closely-spaced array antenna is influenced significantly by the strong mutual interaction between the array elements. Therefore the knowledge of mutual coupling effects is essential in the design of array elements. However, how to accurately predict the mutual coupling in a big array in a retractable way is not yet established. A theoretical formulation for arrays with strong mutual coupling between elements is difficult to analyze since these interactions are typically a near-field phenomenon. In this work, a generic analytical based method to compute the mutual coupling between elements in a big array was developed. Based on the relation between the Active Reflection Coefficients (ARC) and mutual coupling, the mutual coupling coefficients can be extracted from Fourier decomposition of the simulated ARC. This method was proven to be more efficient and practical than the full modal method since only a unit cell is modelled. Effects of mutual coupling between TSA elements on the active impedance matching in tapered slot arrays were then investigated.

Gigabit Powerline Communications: Characterisation of Powerline Channels

By: Shu Xian Chen                                                                                      

Powerline has long been used for delivering electricity to households in the nation, the wide spread installations of powerline network infrastructure make it a very attractive way of solving the "last mile" problem and to provide broadband in-home networking. By eliminating new wiring installation or other wireless "hot spot", every device that plugs into a power outlet can connect to the home network. The idea is to input ultra-wide band signal to the PL to achieve wide bandwidth. However, being designed for very low frequency (50~60Hz) band, powerline is a harsh environment for broadband communication and a number of obstacles need to be tackled to achieve this goal.

Stability Comparison Between Multi-Resolution Time-Domain (MRTD) and Finite-Difference Time-Domain (FDTD) Techniques

By: Xiaojing Wang

The Multi-Resolution Time-Domain (MRTD) technique is a superior numerical algorithm over the long established Finite Difference Time-Domain (FDTD) technique. The advantages of low numerical dispersion and less memory intensive are particularly of interest for analyzing hybrid structures involving antennas and metamaterials.  In this paper, the Haar wavelet based MRTD is reviewed. A stability comparison between MRTD and FDTD is presented in two-dimensional simulations with various configurations involving layered dielectric cylinders. The superiority of MRTD over FDTD is demonstrated.

Full-Wave Finite-Difference Time-Domain Simulation of Electromagnetic Cloaking Structures

By: Yan Zhao

This research proposes a radial dependent dispersive finite-difference time-domain method for the modelling of electromagnetic cloaking structures. The permittivity and permeability of the cloak are mapped to the Drude dispersion model and taken into account in dispersive FDTD simulations. Numerical simulations demonstrate that under ideal conditions, objects placed inside the cloak are `invisible' to external electromagnetic fields. However for the simplified cloak based on linear transformations, the back scattering has a similar level to the case of a PEC cylinder without any cloak, rendering the object still being `visible'. It is also demonstrated numerically that the simplified cloak based on high-order transformations can indeed improve the cloaking performance.

A 31.5GHz Patch Antenna for Medical Implants

By: Yasir Ahmed  

We have proposed a 31.5GHz patch antenna for medical implants. The design is based on the transmission line model and is simulated in CST. The patch antenna performs reasonably well in terms of return loss and radiation efficiency. However, the most attractive feature of this design is its form factor. Typical antennas designed for the microwave range are quite large in size, which makes them unsuitable for implants. The proposed design is much smaller in size but still retains the essential characteristics for reliable communication.

Millimetrewave Metamaterials with Controlled Defects and Their Applications

By: Yoonjae Lee

Millimetrewave systems are becoming increasingly important in many applications because they can provide wider bandwidth for transmitting large amount of data and better resolution in radar systems. Electromagnetic Bandgap (EBG) structures, a class of metamaterial and also known as photonic bandgap structures (PBG) in optics, are now finding numerous applications at microwave and millimetrewave frequencies. The full potential of EBG structures can be utilised with a full three dimensional (3D) bandgap. Thus rapid and cost-effective fabrication techniques for 3D EBG structures are of significant importance. A woodpile structure exhibits a full 3D bandgap and can be easily fabricated for applications at microwave frequencies using columns of individually machined dielectric rods. However, at millimetrewave frequencies, conventional machining would not be convenient because of small dimensions (50–500mm). Sophisticated techniques are available for microstructures, but those are more appropriate for terahertz and photonic wavelength applications, and would be costly to fabricate 3D structures with large number of layers for applications at W-band (75-110 GHz). Recently, we have demonstrated a direct rapid prototyping method for constructing 3D EBG materials for millimetrewave applications, with a possible extension to higher frequencies based on extrusion freeforming of ceramic materials, which can also be versatile for constructing curved geometries and creating defects in layered structures. In this paper, we present the characteristics of freeformed EBG structures with controlled defects and their applications for antennas and components at millimetrewave bands.

Design of a Wideband Thin Helical Antenna for Cancer Treatment

By: Zhenxin Zhu

Recently, it has been demonstrated that ultra-short electrical pulses can be used as a purely electrical cancer therapy to kill tumours without hyperthermia or drugs. For the deep-seated tumours, the interstitial type applicators, i.e. antennas, are required. Reducing the duration of electrical pulses into the sub-nanosecond (100 or so ps) range requires the thin, wideband and near-field-focusing antennas to generate pulsed electric fields with high intensity in tissues. However, the existent microwave antennas for hyperthermia are narrow band. We have presented a numerical investigation of a wideband coaxial-based helical antenna for deep-seated cancer treatment.  The proposed antenna with a diameter of 1.79mm and the total length of 20mm, which is thin and small enough for invasive cancer treatment. Numerical simulations indicate that the antenna exhibits an acceptable return loss (5.6dB) over a wide frequency range from 2.82 to 11 GHz. The electric field distributions are confined along the antenna axial. We are now in the process of testing this type of antenna in experiment to verify its suitability for deep-seated cancer treatment.

2: Centre for Digital Music (C4DM)

Audio Source Separation by Time-Frequency Masking

By: Andrew Nesbit

Audio source separation is an interesting and challenging problem with many practical applications. These include remixing and denoising for generating high quality audio output and information retrieval tasks such as source identification and transcription. Cases in which the number of sources is greater than the number of mixture channels are common, so we use the time-frequency masking framework. Lately we have been focussing on the relationships between audio source separation and audio coding. Our new coding scheme determines highly flexible, (pseudo-)shift-invariant signal transformations which adapt to the signal structures and reduce artifacts at block boundaries, thereby improving separation performance compared to other transforms such as the windowed Fourier transform.

By using oracle estimation techniques, we can determine the most performant signal transformations and then separate the sources by time-frequency masking. This approach to algorithm evaluation requires knowledge of the original source signals and indicates the highest possible potential performance.

Finally, we note that visualisation of the source separation process is useful for framing its applications in an interactive context. We have recently made preliminary extensions to the Audacity audio editing software application to include source separation functionality.

A Real-Time Beat-Tracker for Live Performance

By: Andrew Robertson

Contemporary bands are able to craft intricate sounds in the studio by overdubbing tracks with extra parts and using sequencers and effects. When trying to re-create their sound live, if they wish to have these extra parts played, they must compromise by having the drummer wear headphones or trigger tracks manually. B-Keeper is a beat-tracker that controls the tempo of a sequencer so these parts can be played automatically by the computer which follows the tempo of the band. This way, the band plays live and the computer plays to them, rather than the other way round.

Robustness of Vocal Timbre Features in Live Performance Contexts

By: Dan Stowell

We are developing systems for use in live performances, which analyse and respond to vocal timbre. In order to make such analysis robust, we must identify acoustic features that are resilient against degradations of the source such as crowd noise, feedback echo, or distortion.

In our experiment we applied various simulated degradations to voice signals, and measured the consequent deviation of timbre features from their “clean” values.

We find that general tendencies can be identified - some features are quite generally robust, some are much less so - but also that there are some subtleties to the interaction between features and degradations.

From this work it is possible to identify timbre features that should and should not be used for vocal timbre analysis when robustness is important.

An Automatic Maximum Gain Normalisation Technique with Applications to Audio Mixing

By: Enrique Perez

Live public addressing system that uses a microphone amplifier speaker chain transmits its sound through the air towards the listener. Using the air as a propagation medium has the inevitable effect of turning the sound reinforcement system into an endless feedback loop, and it is the air itself that acts as a feedback path. That is a physical fact and one that cannot be changed, but one we must take into consideration when designing or interacting with a sound system. This is with the aim to reduce audio artifacts due to the feedback path. With this aim in mind, it is the purpose of this presentation to introduce a normalization technique that prevents feedback when interacting with an audio system. The proposed method automates the engineering task of continually revise the system gain structure in order to avoid undesired feedback artifacts. This method permits to achieve maximum gain before feedback while realizing technical constrains off the mixing engineer, permitting him to concentrate more on the aesthetic contributions of the mixture. The method permits the audio mixing engineer to interact with the system without the fear of introducing feedback. The algorithm uses an impulse measurement of a mathematical model of the system to automatically calculate the appropriate gain compensation to avoid undesired artifacts due to feedback.

Semantic Music Analysis for Intelligent Audio Editing

By: George Fazekas

Off-the-shelf sound editors for professional music production do not have formalized knowledge about the edited material, thus human-computer interaction is inefficient. We believe that recent development in music information retrieval and audio analysis is capable of bringing about significant improvement. The primary motivation behind this research is the adaptation and development of music analysis tools for creative music production. The idea rises from the recognition that the fundamental way of editing music, even though it requires high level of human interaction with often cumbersome work, remains unchanged throughout the last decades. The objective is to enhance the workflow and user experience in the post-production environment of contemporary popular music recordings in the broadest sense.

Our practical goal is enhancing sound editing routine through providing visual cues related to the music structure and intelligent context dependent functionality. The research builds on technologies for audio feature extraction and methods for inferring higher level meaning in the perceptual, semantic or musical domain (e.g. note onset detection, perceptual or semantic segmentation). Which of these data proves to be the most relevant in developing an enhanced human interface is an open research question.

Separation of Speech Signals Using a Sparse Dictionary Algorithm

By: Maria Jafari

The problem of source separation in echoic and anechoic environments is addressed with a new algorithm which adaptively learns a set of sparse stereo dictionary elements, which are then clustered to identify the original sources. The atom pairs learned by the algorithm are found to capture information about the direction of arrival of the source signals, which allows to determine the clusters. A similar approach is also used here to extend the dictionary learning K singular value decomposition (K-SVD) algorithm, to address the source separation problem, and results from the two methods are compared. Computer simulations indicate that the proposed adaptive sparse stereo dictionary (ASSD) algorithm yields good performance in both anechoic and echoic environments.

Automatic Annotation and Music Retrieval Using a Joint Aspect Model

By: Mark Levy

We investigate the use of a simple latent aspect model trained jointly on audio features and social tags for a collection of tracks. The model can be used to annotate untagged or partially tagged tracks, to retrieve tracks by free text queries such as “laid back piano jazz”, and for query by example to find nearest neighbouring tracks for music recommendation and playlist generation.

Automatic Rhythm Transformations of Musical Audio

By: Matthew Davies

We present a new method for the rhythmic transformation and syncrhonisation of musical audio signals. Our approach relies on a metrically informed rhythmic segmentation defined in terms of automatically extracted beat locations and bar boundaries. Each rhythmic segment can then be time-scaled to alter the overall rhythmic properties of an input signal while preserving beat and metre. Given two such rhythmic segmentations we can time-scale the rhythmic segments in one music signal to match those of the other. The resulting effect can be considered a novel form of rhythmic synchronisation between two musical signals.

Interactive Auralisation System Using High-Quality Impulse Responses

By: Rebecca Stewart

Convolution reverberation is an excellent method for generating high-quality artificial reverberation that accurately portrays a specific space, but it can only represent the static listener and source positions of the measured impulse response being convolved.  A system is presented here that allows for multiple impulse responses with varying source and receiver configurations are convolved with dry input audio to create the illusion of a moving source. The computational cost is decreased by using a truncated impulse response which contains only the early reflections while the late reverberation is simulated with an averaged impulse response tail.  Only the early reflections change according to the change in position while the same reverberation tail is used throughout. 

Object Coding Audio Using MIDI and DLS

By: Steve Welburn

Object coding offers a general framework in which to reconstruct an audio signal using multiple parametric instrument models. Work to date has concentrated on low bit-rate sound production from a limited number of audio models, principally sinusoidal, harmonic and plucked strings, with some work on noise modelling.

The MIDI standard offers a protocol for controlling musical devices, and within the constraints of this system, we can only approximate original audio, as the standard has little bearing on the sounds produced for those parameters. The MIDI Manufacturer’s Association (MMA) have addressed this in two ways: General MIDI (GM), a standard set of instruments, although the precise sounds are not standardised; and Downloadable Sounds (DLS), a standard for wavetable synthesis, targetted at consistency of audio across platforms.

We are investigating the combination of MIDI and DLS as an object-based representation of audio.

Sinusoid Modelling in a Harmonic Context

By: Wen Xue

This article discusses harmonic sinusoid modeling. Unlike standard sinusoid analyzers, the harmonic sinusoid analyzer keeps close watch on partial harmony from an early stage of modeling, therefore guarantees the harmonic relationship among the sinusoids. The key element in harmonic sinusoid modeling is the harmonic sinusoid particle, which can be found by grouping short-time sinusoids. In-stead of tracking short-time sinusoids, the harmonic tracker operates on harmonic particles directly. To express harmonic partial frequen-cies in a compact and robust form, we have developed an inequalitybased representation with adjustable tolerance on frequency errors and inharmonicity, which is used in both the grouping and tracking stages. Frequency and amplitude continuity criteria are considered for tracking purpose. Numerical simulations are performed on simple synthesized signals.

Interlinking Music Datasets on the Web

By: Yves Raimond

In this poster, we detail how Semantic Web technologies are used to interlink heterogeneous music-related datasets. In particular, we give an overview of what has been achieved by the Linking Open Data on the Semantic Web community project. This project gave birth to a whole new range of applications, consuming linked data from all around the Web including editorial, encyclopedic, geographic or temporal data, as well as content-based data. Such applications include personal music collection management, playlist creation, music recommendation, and data sharing among music information retrieval researchers.

3: Multimedia & Vision Research Group

Unsupervised Trajectory Clustering for Video Analysis

By: Anjum Nadeem

The analysis of moving objects from video has been attracted much attention in the research community in recent years for its impact in applications like video surveillance, traffic monitoring and activity recognition. Moving objects generate trajectory data that grows with time and with the number of objects. For the analysis of these data, more human efforts are required for the understanding and the interpretation of the inherent information. Trajectory clustering categorises raw trajectories in a way that is more manageable and comprehensible for a human operator or for an automated algorithm.
In this work we propose an unsupervised approach for trajectory clustering. The proposed approach is divided into four main steps: first we transform the trajectories into a set of feature spaces and then Mean-shift is used to identify the modes and the corresponding clusters in each space. Furthermore, a merging procedure is devised to refine these results by combining similar adjacent clusters. The final common patterns are estimated by fusing the clustering results across all feature spaces. Clusters corresponding to reoccurring trajectories are considered as normal, whereas sparse trajectories are associated to abnormal and rare events. The performance of the proposed algorithm is evaluated on real outdoor video surveillance scenarios and it is compared with state-ofthe- art techniques.

User Relevance Feedback Using Particle Swarm Optimisation

By: Chandramouli Krishna

In visual information retrieval, due to the existence of "semantic gap", the performance of content based image retrieval has not been satisfactory. Over the last decade, a number of researchers have developed many pattern recognition algorithms for image classification, content retrieval. The performance of the systems have been far from the expectation of the users, due to user's subjectivity attributing to the semantic gap. The most natural way of getting user's subjective information and preferences into the system is by creating models that incorporate online learning from the user interactions with the search engine. The idea underpinning this model is a "relevance feedback" loop with the user at the centre and the machine learning from the user's feedback. In this work, we present the investigations carried out using Particle Swarm Optimization for solving fitness functions in pattern recognition algorithms. The experimental results show an improved performance of the retrieval system, incorporating Particle Swarm Optimization.

Enabling Access to Sound Archives through Integration, Enrichment and Retrieval

By: Damnjanovic Ivan

Many digital sound archives still suffer from tremendous problems concerning access. Materials are often in different formats, with related media in separate collections, and with non-standard, specialist, incomplete or even erroneous metadata. Thus, the end user is unable to discover the full value of the archived material. To expose the inherent value of the archived material, powerful multimedia mining techniques are needed, in combination with content extractors, meaningful descriptors, and visualization tools.

Classification of Visual Information Using Spectral Clustering

By: Damjanovic Uros

As we entered the 21st century, huge amount of available digital information resulted in great research efforts being put in development of systems for efficient organization, storage and delivery of multimedia items to end users. Understanding of the underlying content and its delivery to the user is goal of modern content based systems. Fundamental step towards creation of content based image and video retrieval systems is clustering and classification. Detecting underlying patterns in the multimedia items and organization these items based on actual content is cornerstone for efficient browsing and retrieval of visual information. Spectral clustering emerged last years as promising clustering tool that use eigenvectors and eigenvalues of underlying similarity matrix. Although there have been several proposals to cope with the clustering task by means of eigenvectors, little research has been done on the use of spectral methods for the classification of visual data. We investigate the application of spectral methods for different problems in computer vision such as image classification, shot boundary detection and video summarization.

A Knowledge Structuring Technique for Image Classification

By: Dong Le

A system for image classification based on a knowledge structuring technique is presented. The knowledge structuring technique automatically creates a relevance map from natural images. It also derives a set of well-structured representations from lowlevel description to drive the final classification. Classification is achieved by simulating high-level top-down visual information perception. The proposed modular system architecture offers straightforward expansion to include contextual input, and multimodal information if available.

Image Visualisation

By: Janjusevic Tijana

Image visualization as a research area addresses the problem of presenting the content of large video and image database in intuitive and concise way. In order to efficiently explore large image or video collections, users should be able to have good overview of the database. For this reason visualisation solution should provide enough information to the user in such way that he understands relations and content stored in the database.

There has been a lot of work done regarding development and design of image and video visualization systems, with different image processing tools, browsing and visualization algorithms. Since the result is high user dependent there is always an open issue of the quality and efficiency of the existing visualization systems.

This research aims at introducing novel visualisation approach that will enable user to distinguish dissimilar sets of images or video keyframes displayed and also propose layout method for mapping similarity relations between images/keyframes and cognitive partitioning of display space.

Detection and Tracking of Humans and Faces

By: Karlsson Stefan

We present a video analysis framework that integrates prior knowledge in object tracking to automatically detect humans and faces, and can be used to generate abstract representations of video (key-objects and object trajectories). The analysis framework is based on the fusion of external knowledge, incorporated in a person and in a face classifier, and low-level features, clustered using temporal and spatial segmentation.

Low-level features, namely, color and motion, are used as a reliability measure for the classification. The results of the classification are then integrated into a multitarget tracker based on a particle filter that uses color histograms and a zero-order motion model. The tracker uses efficient initialization and termination rules and updates the object model over time. We evaluate the proposed framework on standard datasets in terms of precision and accuracy of the detection and tracking results, and demonstrate the benefits of the integration of prior knowledge in the tracking process.

Robust Homography-Based Trajectory Transformation for Multi-Camera Scene Analysis

By: Kayumbi Gabin

In this work, we present a robust image registration algorithm for the estimation of extended object trajectories in a distributed camera setting. The registration algorithm is based on a statistical homography estimation that accounts for errors related to the estimation of the homography matrix. Unlike traditional approaches based on Singular Value Decomposition (SVD), we derive a planar homography estimation from the renormalization technique. We demonstrate the proposed algorithm with the generation of the mosaic of the observed scene as well as with the registration of the spatial locations of moving objects (trajectories) from multiple cameras. Finally, we compare the transformed trajectories with those obtained with the homography estimated by SVD and Least Mean Square (LMS) methods and discuss the improvement in terms of spatial accuracy using objective evaluation metrics on standard test sequences.

Objective Evaluation of Pedestrian and Vehicle Tracking on the CLEAR Surveillance Dataset

By: Maggio Emilio

Video object detection and tracking in surveillance scenarios is a difficult task due to several challenges caused by environmental variations, scene dynamics and noise introduced by the CCTV camera itself. In this work, we analyse the performance of an object detector and tracker based on background subtraction followed by a graph matching procedure for data association. The analysis is performed based on the CLEAR dataset. In particular, we discuss a set of solutions to improve the robustness of the detector in case of various types of natural light changes, sensor noise, missing detections and merged objects. The proposed solutions under various parameter settings are analysed and compared based on 1 hour 21 minutes of CCTV surveillance and the CLEAR evaluation metrics.

Region Segmentation and Feature Point Extraction on 3D Faces Using a Point Distribution Model

By: Nair Prathap

We present a novel approach to accurately detect landmarks and segment regions on face meshes without the use of texture, pose or orientation information. The proposed approach is based on a 3D Point Distribution Model (PDM) that is fitted to the region of interest using candidate vertices extracted from low-level feature maps. The robustness of the algorithm is evaluated in the presence of noise and at the variation of the number of scans and model points used in the learning phase. Experimental results demonstrate the accuracy of the proposed method in detecting landmarks, with an improvement of 55% over a state-of-the-art non-statistical approach.

Patch-Based Image Analysis for Classification and Retrieval

By: Passino Giuseppe

The research field of content-based image classification and retrieval (CBIR) is related to the automatic analysis and understanding of digital images. This is a stimulating area both for important speculative aspects related to machine understanding, and for the considerable range of applications. A part-based approach to the problem involves the analysis of images through reasoning on single constituent components (referred as parts or patches). This approach to the problem has shown promising developments and the underlying idea is also sustained by the analysis of biological vision systems. However, the inherent drawback of such a method is the exponential growth of the complexity of the problem, when compared to standard global image analysis systems.
Tenacious efforts have been made to design frameworks able to handle the image analysis problem by exploiting parts relationships while retaining the complexity to an affordable level. This research focuses on how to embed the parts relationships in a probabilistic framework based on graphical models. In particular, discriminative models such Conditional Random Fields are used for the analysis of extracted image parts. Both problems of patches location and identification, and problems related to inference are studied, and novel solutions are proposed.

Semantic Multimedia Retrieval using Ant Colony Inspired Methods

By: Piatrik Tomas

A fundamental step towards content based image and video annotation and retrieval is classification or clustering. Although there have been several proposals to cope with the underlying classification task, little research has been done on the use of bio-inspired algorithms for the classification of visual data. We investigate the application of algorithms based on the behavior of real ants for image classification and video summarization. The Ant Colony Optimization (ACO) learning mechanism is integrated into a COP-K-means approach to solve image classification problems. In the second proposal, we model the ability of ants to build live structures with their bodies in order to discover, in a distributed and unsupervised way, a tree-structured organization and summarization of the video data.

Region-of-Interest Video Encoding Using Object Tracking Data

By: Popkin Timothy

Region-of-interest video coding works by varying the bitrates of different parts of the scene according to their visual importance (saliency). We present some results of regionof- interest encoding using saliency information taken from the output of a recentlyintroduced object tracking technique that constists of a statistical change detector with graph-matching-based object tracking. Two approaches were employed for encoding video using the saliency information. Firstly, selective pre-blurring was performed in conjunction with an ordinary MPEG-2 encoder. Secondly, the same pre-blurring was
performed in conjunction with an MPEG-2 encoder of which the quantization stage had been modified so as to increase the bitrate savings due to run-length coding. For fixed bitrate MPEG-2 the modified encoder was found to produce video of poorer visual quality than through pre-blurring alone. Additionally, we propose a technique for the efficient computation of selective Gaussian blurring.

Optimal Joint Source Channel Coding for Scalable Video Transmission Over Wireless Channels

By: Ramzan Naeem

A robust and novel approach for optimal bit allocation between source and channel coding is proposed. The proposed approach consists of a wavelet-based scalable video coding framework and a forward error correction method based on the serial concatenation of LDPC codes and turbo codes. Turbo codes shows good performance at low signal to noise ratios but LDPC outperforms turbo codes at high signal to noise ratios. So the concatenation of LDPC and TC enhances the performance at both low and high signal to noise ratios. The scheme reduces the video distortion at the decoder under band-with constraints. The reduction is achieved by efficiently protecting the different quality layers from channel errors. Furthermore, an efficient decoding algorithm is proposed that reduces the decoding complexity of channel decoder. Experimental results clearly show that the proposed approach outperforms conventional forward error correction techniques

Multi-Camera Scene Analysis Using an Object-Centric Continuous Distribution Hidden Markov Model

By: Taj Murtaza

We propose a multi-camera event detection framework that can operate on a common ground plane as well as on the image plane. The proposed event detector is based on an object-centric state modeling that uses a Continuous Distribution Hidden Markov Model (CDHMM). Video objects are first detected using statistical change detection and then tracked using graph matching. Next, the algorithm recognizes events by estimating the most likely object state sequence using a HMM decoding strategy, based on the Viterbi algorithm. We demonstrate and evaluate the proposed framework on standard event detection datasets with single and multiple cameras, with both overlapping and nonoverlapping fields of view.

Video Event Segmentation and Visualisation in Non-linear Subspace

By: Tziakos Ioannis

We introduce the use of dimensionality reduction for video event detection without explicitly using motion estimation or object tracking. Raw data from video sequences are used to construct a low dimensional mapping representing the input frames. We compare Principal Component Analysis, Multidimensional Scaling, Isomap, Maximum Variance Unfolding and Laplacian Eigenmaps and implement an approach
based on local, non-linear dimensionality reduction. We propose an approach with a graph based on the similarity of frames and enriched with the temporal information from the sequence processed by Laplacian Eigenmaps. This makes it possible to visualise the manifold of motion in the scene and to detect unusual events in a low dimensional space. We demonstrate the approach on standard traffic surveillance test sequences.

Optimised Compression Strategy in Wavelet-Based Video Coding Sing Improved Context Models

By: Zgaljic Toni

Accurate probability estimation is a key to efficient compression in entropy coding phase of state-of-the-art video coding systems. Probability estimation can be enhanced if contexts in which symbols occur are used during the probability estimation phase. However, these contexts have to be carefully designed in order to avoid negative effects. Methods that use tree structures to model contexts of various syntax elements have been proven efficient in image and video coding. In this paper we use such structure to build optimised contexts for application in scalable wavelet-based video coding. With he proposed approach context are designed separately for intra-coded frames and motioncompensated frames considering varying statistics across different spatio-temporal subbands. Moreover, contexts are separately designed for different bit-planes. Comparison with compression using fixed contexts from Embedded ZeroBlock Coding
EZBC) has been performed showing improvements when context modelling on tree structures is applied.

Unequal Error Protection in Mobile Channel Transmission Using H.264/AVC

By: Tan Keyu

Video compression has become an essential technology for multimedia technologies such as broadcast and entertainment media. However, delivery of video is still a challenging problem today especially for mobile transmission because of the commonly used error prone transmission channels such as wireless networks and terrestrial broadcasting. The data loss during the transmission may cause unacceptable visual quality degradation at the decoder. Unequal Error Protection (UEP) is one of the crucial error resilient encoding approaches for error robustness. As the most developed video coding standard, H.264/AVC has applied efficient error resilience tools that could be used for UEP, such as Flexible Macroblock Ordering (FMO). Research has been done to make use of the flexibility of these tools. In the previous work of Marcus Im, a new MB classification algorithm was proposed for choosing slice groups. This method can greatly improve the coding performance. The proposed research is based on this to use MB importance identification with Region of Interest.

4: Networks Research Group

Intelligent Resource Management in Next Generation Optical Networks

By: Ali Hassan

All-optical D-WDM networking is a promising technology for next generation optical networks (NGON). One of the desirable features of NGON is fast and automatic service provisioning. Traditionally, global-optimum search schemes like integer linear programming (ILP) are used to find an optimal solution. These techniques are unusable for connection provisioning in dynamic environment, because of the computational complexity and execution time. Different heuristic and stochastic techniques that are used for dynamic provisioning can lead the network configuration away from optimal/near-optimal configuration. Therefore, in dynamically provisioned optical networks, there exists a strong need to employ re-optimization mechanisms so that the state/configuration of the network is kept optimal. Re optimization is an expensive function of the management plane, and should only be used if the re optimization can bring some significant saving in terms of number of resources. In this research, intelligent hybrid architecture has been proposed for resource management in next generation optical networks using particle swarm optimization. The architecture also employs pro-active resource management for subsequent connection requests using traffic prediction information so that minimal re-deployments of lightpaths are required to bring the network back to an optimal / near-optimal configuration.

Intelligent Management of Context-Aware Services

By: Dejian Meng

This research proposes a context-aware framework including three interlinked models: a context composition model for querying and composing contexts, a semantic mediation model to resolve various context heterogeneities, and a context-aware negotiation model to achieve commongoal situation in context-aware systems. This framework is applied to the NVT (New Visitor Tour) scenario, where different situated contexts are used to adapt an individual user's activities and to adapt multi-party situated activities such as an ad-hoc scheduled meeting. A context-aware system has been designed that includes context management, context-service mediation, context interoperability and various context-aware services such as location-aware services linked to maps and personalisation.

Queueing Network Topology for Modelling Cellular/Wireless LAN Interworking Systems

By: Guozhi Song

A network of queues (or queueing network) is a powerful tool in the performance evaluation of many complex systems. Obtaining the network topology of a system is usually the very first step to construct a good queueing network model. In this research, we model an integrated system of cellular and wireless local area networks (WLANs) as an open network of Erlang loss systems. A new network topology representation of the system is also proposed to facilitate the analysis of handover traffic rates.

The performance measures of blocking probabilities and dropping probabilities are calculated using the Erlang fixed-point method. A simple scenario of three cellular cells and three WLANs interworking is set up and the numerical results show a good agreement with the realistic situation.

Sustaining Wireless Coverage in Cellular Networks in Disaster and Emergency Scenarios

By: Haris Pervaiz

In disaster scenarios most of the local communication infrastructure would become unavailable. The aim is to investigate about the remaining existing communication infrastructures and come up with the proposed scheme (either terrestrial or hybrid technology based) to establish fast and rapidly deployable wireless emergency communication. The satellites can only be used for backhaul purposes.
To develop a framework to maintain and optimize wireless communications in a crisis situation, by dealing with the main elements that can affect or lead to communication system failures i.e. loss of available infrastructure, lack of available deployed infrastructure, overload on presently available infrastructure and difficulties associated with the rapid deployment and integration of new temporary infrastructure.
The concept of Recovery level is defined and will be simulated for different types of wireless networks. The findings will be compared against the measurable outcomes like retention of intra-cell communication in case of backhaul failure, extension of communication capability to neighborhood communications using remaining infrastructure and creation of new generic service environment supporting adaptive emergency services for disasters. This work will analyze different emergency scenarios from resource management perspective and suggest their specific solutions.

Ubiquitous Knowledge-Based Management of ICT Services in the Home

By: Ioannis Barakos

Activities in an ICT home include the operation of devices and services (computing devices, audio/video equipment, house appliances, sensors) throughout the home and the dynamic installation, maintenance and termination of these devices and services. This research project is focusing in two types of interaction in an ICT home environment: Computer Devices to Computer Device Interaction (CCD) and Computer Device to Human Interaction (CHI). Furthermore, there are three common strategies to manage the home services: the Do-It-Yourself approach where the home occupant manage a service using the user-manual's instructions, the out-sourced system management approach where the home occupant asks for commercial engineer support and the automatic management approach where the service supports a mechanism to automatically configure and manage itself. The main objective of this research project is to design, investigate and validate models that support the management of future ICT home services, focusing in particular on CHI and CCI models. The strategy to achieve this is to develop an initial framework that supports CCI interaction. Then another framework will be developed to support CHI interaction. Later these two frameworks can be combined to one including both CCI and CHI interaction models to support the three management approaches described above.

Cooperative & Adaptive Radio Coverage in Cellular Communication Networks

By: Jiayi Wu

The research work presented here is about an innovative approach that improves radio resource efficiency of cellular networks. The suggested scheme aims to introduce flexibility into conventional style of fixed cellular coverage, which can have problem reaching satisfactory service probability when the cellular topology and its resource map are getting unfit with time-varying geographic traffic distribution.

Combining the strength from both radio resource managment and antenna array technologies, the new scheme tackles the resource allocation problem with scenario fitness provided by cross-cell load balancing. To achieve this, the frontier of cells in a network are adapted to better divide and equalise the demand between them. Cooperation is used for finding ideal solutions of dividing and dedicated algorithms are designed to facilitate quick resolution of problems.

Ubiquitous, Flexible and Distributed Computing

By: Jun Zhang

Motivated by the observation that there exists a lot of cheap and networked computing power and the inflexibility of Transmission Control Protocol (TCP) in the context of mobility, this research proposes a new system enabling applications to roam seamlessly while they are actively processing executable code and/or communicating with remote entities. This architecture effectively decouples the running of applications from the underlying environment, facilitating both security and opportunistic processing on transitory resources. This decoupling also promotes the ability to create multiple clones of an application for performance gains. To improve the performance of the system, Fuzzy logic and Case Based Reasoning (CBR) have been applied to the system to better regulate activities such as job migration between resources or client connection handover from one Proxy to another.

Traffic Congestion Forecasting in WCDMA Networks

By: Kejing Zhang

Congestion control is considered as a key factor in radio resource management. Early detection and rapid resolution network congestion can considerably improve network capacity. Increasing numbers of users and high bit-rate services complicate radio resource management in 3G systems. As the same set of frequencies are used in each cell in CDMA systems, the solutions of traffic load balancing in 2G systems, which are based on using different frequency channel assignment, are no longer applicable.

One viable approach is to use smart antenna techniques to dynamically change the radiation pattern of target cells to reduce congestion. This research introduces a traffic forecasting method based on maximum likelihood estimation, which is able to predict traffic patterns when there is random movement of users. Then, the radiation patterns of smart antennas can be coordinated according to the prediction results to increase the network capacity.

Cross-Linked Multi-Semantic Multi-Mode Annotations for Images Retrieval

By: Kraisak Kesorn

Almost all documents on the Internet are now usually composed of several kinds of multimedia information for personal, entertainment, and scientific purposes. These kinds of multimedia require multimedia retrieval systems, rather than traditional text information retrieval system. While numerous ways of performing this task have been conceptualised and implemented, the area is still very much in its infancy, and most of the associated problems remain unsolved. Nowadays, there are huge amounts of images produced every day for news, sport, entertainment, and education by media companies and various institutions. Among these photo categories, the sport domain is receiving growing attention due to the interests of publishers, broadcasters, producers, sponsors, and audience worldwide. Internet users are very demanding while searching for photographic information about, for example, events (e.g. particular goal), athletes, sport types, and places. Therefore, the need of the semantic annotation system for sport images is vital to cater for the user's needs. The central theme of this research is to improve image retrieval using semantics for sport domain. Knowledge base technique will be adopted into this research in order to attempt to extract knowledge from these multimedia documents and, and used as knowledge for semantic searching and supporting multiinterpretation for a single instance content from different users. In addition, surrounding text information of the image will be cross-linked to visual features to enhance the image annotation system.

QoS Based Vertical Handoff Between WLAN and WiMAX Compatible with the 802.21 Framework

By: Lina Men

This research aims to maximise the network throughput of an integrated wireless network of WLAN and WiMAX while maintain acceptable QoS (packet loss and packet delay etc.) of applications, through the approach of designing QoS based Vertical Handoff (VHO) triggers and QoS based VHO decision making algorithms (DMA).

In an all IP based, hybrid wireless network environment, using QoS based VHO trigger and QoS based VHO DMA can effectively utilise the complementary characteristics of WLAN and WiMAX, especially their different QoS support due to their different MAC layer. “QoS based” indicates VHO triggers and VHO DMA metrics include QoS performance parameters, such as, packet loss and packet delay of a certain on-going application, traffic load and access delay of a WLAN network, network utilization of a WiMAX network etc.

Simplification of Large Complex Networks via Cycle Conservation

By: Ling Liu

Modelling the fine granularity dynamics in large complex networks (e.g. Internet) is difficult due to the sheer size of the network. Hence the analysis and modelling can be done in smaller networks which conserve the relevant topological properties. As in many networks the resilience of the network is related to path diversity, here we present a method to simplify large networks by conserving the number of alternative paths or cycles.

The simplification from the original network to a smaller network is done by node contraction. We refer the procedure of contracting all the nodes of the network as deforestation and the contracted network as the skeleton network. The nodes of the skeleton network are a mixture of super-nodes and original nodes. The number of nodes and links, and their connectivity are not unique in the skeleton network; they depend on the order in which the nodes were contracted. The skeleton network consists of only cycles, the majority short cycles, from which the cycle basis can be extracted. And also the topological characteristics of the skeleton networks have been studied.

Novel QoS in Node-Disjoint Routing for Ad Hoc Networks

By: Luo Liu

Ad hoc network (MANET) is a collection of mobile nodes that can communicate with each other without using any fixed infrastructure. To support multimedia applications MANETs require an efficient routing protocol and quality of service (QoS) mechanism. Node-Disjoint Multipath Routing Protocol (NDMR) is a practical protocol in MANETs: it reduces routing overhead dramatically and achieves multiple node disjoint routing paths. QoS support in MANETs is an important issue as besteffort routing is not efficient for supporting multimedia applications. This paper presents a novel adaptation of NDMR: QoS enabled NDMR, which provides QoS support for MANETS. This paper also introduces the limitation of NDMR and shows the performance comparisons.

Monitoring the Aged in a Home Environment - Analysis of the Intentions of Elderly People with Alzheimer’s Disease while Conducting Everyday Activities of Daily Life (ADL)

By: Naeem Usman

In today’s working world the elderly are often classified as a set of dependent people and are sometimes neglected by society. Statistically, after toddlers it is the elderly who are observed to have higher accident rates while performing everyday activities. Alzheimer’s disease is also one of the major impairments that elderly people suffer from, and therefore this leads to the elderly person not being able to live an independent lifestyle due to forgetfulness. One of the ways that will allow elderly people to live an independent life as well determining whether they are safe in their home is to find out what activities the elderly person is carrying out at a given time and provide appropriate assistance or institute safeguards. A two tiered approach is adopted, where a set of common tasks associated with activities in daily life are modelled using a plan representation language. The recognition of the activities is based on recognising plans from constituent tasks, while the tasks are recognised from sensor event sequences based on everyday objects used at home that are captured by simple non intrusive sensors and RFID tags. Specifically there is no use of any cameras or visual surveillance equipment.

Internet Telephony in Multi-Scale 802.11 Wireless Networks

By: Oliver Shepherd

802.11 wireless networks are becoming the method of choice for providing client access to the Internet. These networks are being implemented on many scales, from small scale home solutions to commercial ‘hot spots’ and more recently metropolitan areas, providing blanket coverage, over kilometres. 802.11 is a contention based protocol and more stations sharing the medium result in longer delays. This issue can be a significant problem for real-time traffic such as voice; a highly time critical service.

In this presentation we explore how voice traffic is affected when using these wireless contention based protocols. To tackle the challenges, we use a Quality of Experience measure to show if a call is acceptable. In the large scale, multiple mediums overlap and the challenge is to reduce these occurrences by optimising channel and power.
We propose a framework to show how the proportion of acceptable calls varies over different optimisation schemes and the trade-offs, such as coverage, this involves. Based on a realistic metropolitan area, simulated on high performance computational resource, initial results are shown using the framework.
Future analysis will bring together the small and large scale explorations with the aim of estimating large scale performance using new analytical models.

Self Organizing Relay Stations in Relay Based Cellular Networks

By: Peng Jiang

The increasing popularity of wireless communications and the higher data requirements of new types of service lead to higher demands on wireless networks. Relay based cellular networks have been seen as an effective way to meet users’ increased bit rate requirements while still retaining the benefits of a cellular structure. The objective of my research is to illustrate effective and efficient approaches for fast deployment of relay stations into existing cellular networks and to demonstrate improvements to system capacity and performance resulting from cooperation between network components. Cooperative control based on geographic load balancing is employed to provide flexibility for the wireless network and respond to changes in the environment. Flexibility in the antenna system is used to provide coverage at the right place at the right time. Experiments show that the proposed approach has significant improvements and robustness in heterogeneous traffic scenarios, even in disaster conditions where some base stations fail. The performance evaluations are based on Mobile WiMAX technology, but the concepts apply more generally.

Intelligent Radio Access Selection on Heterogeneous Wireless Networks

By: Weizhi Luo

Some of the important advantages of having an integrated mobile network environment would be seamless communications, joint resource management and adaptive quality of service. In such environment, operators would not need to reject the user requests, but redirect them to appropriate networks. However, the sought aims of an integrated network system still have many pending issues. One of them is the selection of the most appropriate radio access network (RAN) according to the requested service and the context information about the user and the networks. We aim to develop efficient RAN selection algorithms inside a realistic internetworking system architecture. Also, as part of our research, we developed models for UMTS and WiFi networks which facilitate the evaluation of the network resource availability in these networks, which provide a morerealistic input of network context information into RAN selection algorithms. The simulation results show that our RAN selection algorithms can improve the network performance.

The DVM Based Dynamic VPN

By: Yiran Gao

This project presents a new dynamic Virtual Private Network (VPN) architecture based on a new coordinating component called the Dynamic VPN Manager (DVM). The DVM manages available network resources and allocates these resources to the users’ applications. To achieve this, a resource scheduling algorithm is designed to assign the network resources in the dynamic environments. Simulations are carried to study the proposed algorithm’s performance. The results show that the scheme is feasible and performs well in dynamic network scenarios, where no knowledge of future events exist, compared with static algorithms which are given complete knowledge of all job arrivals.

Multi-Agent and Autonomic System Models of Ambient Intelligence

By: Zekeng Liang

Ambient Intelligence (AmI) is a vision for ubiquitous computing in which digital devices are embedded in a physical environment, and exhibit a strong and explicit notion of person awareness, i.e., devices which recognise people, their situational context and be able to adapt and anticipate peoples’ desires. This alleviates humans from having to interact with systems at a detailed and syntactic level. In order for systems to be person aware, it is proposed that they need to be designed to exhibit a notion of social intelligence in terms of autonomy, proactiveness and shared goals. Such a notion of intelligence is a central notion of multi-agent systems. A first step is to design AmI devices to exhibit social intelligence at an individual and local level, e.g., an AmI light controller can detect if it is dark and a person is present and will switch on and remain on while the person is present. However, such a simple model is limited because the social intelligence is directed at the person. An extension to the ambient intelligence model is needed such that allows it to operate between multiple devices, enabling them to cooperate as a team to support person-awareness, e.g., the social interaction between multiple lights could help to decide which of the lights was redundant and not switch on. This research proposes a model of the social intelligence that allows devices to operate between them. An attempt is given to validate the model by comparing the test results of the three simple simulators (manual control, semi-auto control and self-managing).

An Invariant in Catastrophic Networks

By: Zhijia Huang

Complex networks such as the Internet, power transmission lines and telephone systems are susceptible to catastrophic failures in which the whole network ceases to function. An avalanche-like breakdown is the most common cause of a catastrophe. Failure of a single node, that is sensitive to overloading, would induce load redistribution among the network. This redistribution of load will overload other nodes, triggering an avalanche of failures. As a result, the entire network fragments into disconnected sub-networks. We are interested to know if a catastrophic-network has a distinctive topology and if the cause of the cascading behaviour scales with the size of the network. To be able to address these questions we need to construct catastrophic networks in a controlled and efficient way.