The recent years have witnessed the prevalence and abundance of vision sensors in various applications such as security surveillance, healthcare and Ambient Assisted Living (AAL) among others. This is so as to realize intelligent environments which are capable of detecting users’ actions and gestures so that the needed services can be provided automatically and instantly to maximize user comfort and safety as well as to minimize energy. However, it is very challenging to automatically detect important events and human behaviour from vision sensors and summarize them in real time. This is due to the massive data sizes related to video analysis applications and the high level of uncertainties associated with the real world unstructured environments occupied by various users. Machine vision based systems can help detect and summarize important information which cannot be detected by any other sensor; for example, how much water a candidate drank and whether or not they had something to eat. However, conventional non-fuzzy based methods are not robust enough to recognize the various complex types of behaviour in AAL applications. Fuzzy logic system (FLS) is an established field of research to robustly handle uncertainties in complicated real-world problems. In this thesis, we will present a general recognition and classification framework based on fuzzy logic systems which allows for behaviour recognition and event summarisation using 2D/3D video sensors in AAL applications. I started by investigating the use of 2D CCTV camera based system where I proposed and developed novel IT2FLS-based methods for silhouette extraction and 2D behaviour recognition which outperform the traditional on the publicly available Weizmann human action dataset. I will also present a novel system based on 3D RGB-D vision sensors and Interval Type-2 Fuzzy Logic based Systems (IT2FLSs) ) generated by the Big Bang Big Crunch (BB-BC) algorithm for the real time automatic detection and summarization of important events and human behaviour. I will present several real world experiments which were conducted for AAL related behaviour with various users. It will be shown that the proposed BB-BC IT2FLSs outperforms its Type-1 FLSs (T1FLSs) counterpart as well as other conventional non-fuzzy methods, and that performance improvement rises when the number of subjects increases. It will be shown that by utilizing the recognized output activity together with relevant event descriptions (such as video data, timestamp, location and user identification) detailed events are efficiently summarized and stored in our back-end SQL event database, which provides services including event searching, activity retrieval and high-definition video playback to the front-end user interfaces.

Existing 3D human face modelling methods are confronted with difficulties in applying flexible control over all facial features and generating a great number of different face models. The gap between the existing methods and the requirements of emotional bio-robots applications urges the creation of a generic 3D human face model. This thesis focuses on proposing and developing two new methods involved in the research of emotional bio-robots: face detection in complex background images based on skin colour model and establishment of a generic 3D human face model based on NURBS. The contributions of this thesis are: A new skin colour based face detection method has been proposed and developed. The new method consists of skin colour model for skin regions detection and geometric rules for distinguishing faces from detected regions. By comparing to other previous methods, the new method achieved better results of detection rate of 86.15% and detection speed of 0.4-1.2 seconds without any training datasets. A generic 3D human face modelling method is proposed and developed. This generic parametric face model has the abilities of flexible control over all facial features and generating various face models for different applications. It includes: The segmentation of a human face of 21 surface features. These surfaces have 34 boundary curves. This feature-based segmentation enables the independent manipulation of different geometrical regions of human face. The NURBS curve face model and NURBS surface face model. These two models are built up based on cubic NURBS reverse computation. The elements of the curve model and surface model can be manipulated to change the appearances of the models by their parameters which are obtained by NURBS reverse computation. A new 3D human face modelling method has been proposed and implemented based on bi-cubic NURBS through analysing the characteristic features and boundary conditions of NURBS techniques. This model can be manipulated through control points on the NURBS facial features to build any specific face models for any kind of appearances and to simulate dynamic facial expressions for various applications such as emotional bio-robots, aesthetic surgery, films and games, and crime investigation and prevention, etc.

Technological advances have provided us with the capability to build large intelligent systems capable of using knowledge, which relies on being able to represent the knowledge in a way that machines can process and interpret. This is achieved by using ontologies; that is logical theories that capture the knowledge of a domain. It is widely accepted that ontology development is a non-trivial task and can be expedited through the reuse of existing ontologies. However, it is likely that the developer would only require a part of the original ontology; obtaining this part is the purpose of ontology modularization. In this thesis a graph traversal based technique for performing ontology module extraction is presented. We present an extensive evaluation of the various ontology modularization techniques in the literature; including a proposal for an entropy inspired measure. A task-based evaluation is included, which demonstrates that traversal based ontology module extraction techniques have comparable performance to the logical based techniques. Agents, autonomous software components, use ontologies in complex systems; with each agent having its own, possibly different, ontology. In such systems agents need to communicate and successful communication relies on the agents ability to reach an agreement on the terms they will use to communicate. Ontology modularization allows the agents to agree on only those terms relevant to the purpose of the communication. Thus, this thesis presents a novel application of ontology modularization as a space reduction mechanism for the dynamic selection of ontology alignments in multi-agent systems. The evaluation of this novel application shows that ontology modularization can reduce the search space without adversely affecting the quality of the agreed ontology alignment.

Will it be possible to create a self-aware and reasoning entity that has the capacity for decision making similar to that we ascribe to human beings? Modern agent systems, although used today in various applications wherever intelligence is required, are not ready for applications where human rationalities are usually the only option in making important decisions in critical or sensitive situations. This thesis is a contribution to this area: a decision-making methodology is introduced to address the different characteristics that an agent should have in order to be better trusted with such critical decisions. The work begins with a study of philosophy in the literature (Chapter 2), which reveals that trust is based on emotions and faith in performance. The study concludes that a trustworthy decision has five main elements: it considers options and their likely effects; it predicts how the environment and other agents will react to decisions; it accounts for short- and long-term goals through planning; it accounts for uncertainties and working with incomplete information; and, finally, it considers emotional factors and their effects. The first four elements address decision making as a product of "beliefs"; the last addresses it as a product of "emotions". A complete discussion of these elements is provided in Section 2.1. This thesis is divided into two main parts: the first treats trust as a product of beliefs and the second treats trust as a product of emotions. The first part builds the decision-making methodology based on argumentation through a five-step approach where first the problem situation representing the actions available to the agent and their likely consequences is formulated. Next, arguments to perform these actions are constructed by instantiating an argumentation scheme designed to justify actions in terms of the values and goals they promote. These arguments are then subjected to a series of critical questions to identify possible counter arguments so that all the options and their weaknesses have been identified. Preferences are accommodated by organising the resulting arguments into an Argumentation Framework (we use Value-Based Argumentation [VAF] for this approach). Arguments acceptable to the agents will be identified through the ranking of the agent's values, which may differ from agent to agent. In the second part (Chapters 5 and 6), this methodology is extended to account for emotions. Emotions are generated based on whether other agents relevant to the situation support or frustrate the agent's goals and values; the emotional attitude toward the other agents then influences the ranking of the agent's values and, hence, influences the decision. In Chapters 4 and 6, the methodology is illustrated through an example study. This example has been implemented and tested on a software program. The experimental data and some screen shots are also given in the appendix.

This thesis describes research conducted in the field of image mining especially volumetric image mining. The study investigates volumetric representation techniques based on hierarchical spatial decomposition to classify three-dimensional (3D) images. The aim of this study was to investigate the effectiveness of using hierarchical spatial decomposition coupled with regional homogeneity in the context of volumetric data representation. The proposed methods involve the following: (i) decomposition, (ii) representation, (iii) single feature vector generation and (iv) classifier generation. In the decomposition step, a given image (volume) is recursively decomposed until either homogeneous regions or a predefined maximum level are reached. For measuring the regional homogeneity, different critical functions are proposed. These critical functions are based on histograms of a given region. Once the image is decomposed, two representation methods are proposed: (i) to represent the decomposition using regions identified in the decomposition (region-based) or (ii) to represent the entire decomposition (whole image-based). The first method is based on individual regions, whereby each decomposed sub-volume (region) is represented in terms of different statistical and histogram-based techniques. Feature vector generation techniques are used to convert the set of feature vectors for each sub-volume into a single feature vector. In the whole image-based representation method, a tree is used to represent each image. Each node in the tree represents a region (sub-volume) using a single value and each edge describes the difference between the node and its parent node. A frequent sub-tree mining technique was adapted to identified a set of frequent sub-graphs. Selected sub-graphs are then used to build a feature vector for each image. In both cases, a standard classifier generator is applied, to the generated feature vectors, to model and predict the class of each image. Evaluation was conducted with respect to retinal optical coherence tomography images in terms of identifying Age-related Macular Degeneration (AMD). Two types of evaluation were used: (i) classification performance evaluation and (ii) statistical significance testing using ANalysis Of VAriance (ANOVA). The evaluation revealed that the proposed methods were effective for classifying 3D retinal images. It is consequently argued that the approaches are generic.

One of the key questions facing governments around the world is that of how to increase and maintain the engagement of citizens in democratic processes. Recent thought, both within academia and government itself, has turned to the use of modern computational technology to provide citizens with access to democratic processes. Access to computer and Internet technology by the general public has vastly increased over the past decade, and this wide access is one of a number of motivations behind research into the provision of democratic tasks and processes online. The particular democratic process that forms the focus of this thesis is that of online opinion gathering in order to aid government decision making. The provision of mechanisms to gather and analyse public opinion is important to any government which claims to promote a fair and equal democracy, as decisions should be made in consideration of the views and opinions of the citizens of such a democracy. The work that comprises this thesis is motivated by existing research into harvesting opinion through a variety of online methods. The software tools available largely fall into one of two categories: Those which are not based on formal structure, and those which are based on an underlying formal model of argument. The work presented in this thesis aims to overcome the shortfalls inherent to both of these categories of tool in order to realise a software suite to support both the process of opinion gathering, and analysis of the resulting data. This is achieved through the implementation of computational models of argument from the research area of argumentation, with special consideration as to how these models can be used in implemented systems in a manner that allows laypersons to interact with them effectively. A particular model of argument which supports the process of practical reasoning is implemented in a web-based computer system, thus allowing for the collection of structured arguments which are later analysed according to formal models of argument visualisation and evaluation. The theories underlying the system are extended in order to allow for added expressivity, thus providing a mechanism for more life-like argument within a system which supports comprehensive computational analysis. Ultimately, the contributions of this thesis are a functional system to support an important part of the democratic process, and an investigation into how the underlying theories can be built upon and extended in order to promote expressive argumentation.

This dissertation is devoted to a comprehensive theoretical and modelling study of dynamic modulation characteristics of semiconductor wide-wavelength tunable laser diodes (TLDs). The two major goals were to investigate how modulation properties of a TLD depend on the wavelength tuning, and how the modulation performance of a TLD can be enhanced in terms of the achievable speed and improved frequency chirping behaviour under the direct modulation regimes using external light for optical injection-locking (OIL). It is demonstrated that modulation performance of free running (FR) widely tunable lasers strongly depends on the tuned lasing wavelength. The relaxation oscillation frequency (ROF) of FR TLD increases from 2.2 GHz to 5.5 GHz with tuning. The main results of investigation of modulation dynamics of OIL TLDs include demonstration of substantial (up to an order of magnitude) increase of the ROF and the modulation bandwidth in comparison with the FR regime and investigation of dependence of ROF on the wavelength tuning. The ROF increases to 24 GHz. We prove that the ROF of the OIL TLD is defined by the difference between the injected master laser’s light frequency and the cavity shifted mode frequency. The latter non-lasing mode has been identified as corresponding to the amplified spontaneous emission and was clearly reproduced in CW spectra of a steady-state OIL TLD. This finding has important practical implications as it allows to directly relating the CW lasing spectra fine features with dynamic performance of the OIL TLDs. Important results were obtained for case of large-signal modulation of the OIL TLD and for a large frequency detuning for side-mode optical injection regime when the master laser’s light is injected near the side-mode of FR TLD with large SMSR. Direct large-signal modulation of the OIL TLD using pseudo-random bit sequence shows superior performance in terms of enhanced modulation speed.

Almost half of UK firms claim to have been subject to some sort of cyber-attack or breach in the last 12 months, with an average cost per incident being around £20,000. Yet, even in the face of these ever-mounting threats, cyber security is still treated as an afterthought throughout the systems development lifecycle (SDLC). Though literature is aiming to rectify this mindset through the proposal of multiple software security solutions, there is still a noticeable absence of any usable, expressive tool for designing cyber security into a system at the requirements stages of the SDLC. By not practicing secure by design, there is a risk of: poor defences, confused developers with no security guidelines to work from, a potential redesign of core functionality and very expensive patch management. There have been several attempts at producing a solution, with modelling languages presenting themselves as the perfect platform to specify such designs. One can observe multiple publications throughout literature which propose the extension of these languages to include security expression. However, the ability of these propositions to provide comprehensive expression of the cyber security domain and remain usable alongside their parent modelling language, remains an elusive endeavour. The aim of this thesis is to produce a solution which ensures the practicability of expressive and usable secure by design tool implementation. That is, by conducting an evaluation of existing attempts at security extension and extracting heuristics based on their current failings, combine them with proven scientific principles to produce a framework which will act as its own form of methodology to guide the development of a security extension to modelling languages.

Many systems in the world can be represented as models of complex networks and subsequently be analysed fruitfully. One fundamental property of the real-world networks is that they usually exhibit inhomogeneity in which the network tends to organise according to an underlying modular structure, commonly referred to as community structure or clustering. Analysing such communities in large networks can help people better understand the structural makeup of the networks. For example, it can be used in mobile ad-hoc and sensor networks to improve the energy consumption and communication tasks. Thus, community detection in networks has become an important research area within many application fields such as computer science, physical sciences, mathematics and biology. Driven by the recent emergence of big data, clustering of real-world networks using traditional methods and algorithms is almost impossible to be processed in a single machine. The existing methods are limited by their computational requirements and most of them cannot be directly parallelised. Furthermore, in many cases the data set is very big and does not fit into the main memory of a single machine, therefore needs to be distributed among several machines. The main topic of this thesis is about network community detection within these big data networks. More specifically, in this thesis, a novel approach, namely Decentralized Iterative Community Clustering Approach (DICCA) for clustering large and undirected networks is introduced. An important property of this approach is its ability to cluster the entire network without the global knowledge of the network topology. Moreover, an extension of the DICCA called Parallel Decentralized Iterative Community Clustering approach (PDICCA) is proposed for efficiently processing data distributed across several machines. PDICCA is based on MapReduce computing platform to work efficiently in distributed and parallel fashion. In addition, the real-world networks are usually noisy and imperfect with missing and false edges. These imperfections are often difficult to eliminate and highly affect the quality and accuracy of conventional methods used to find the community structure in the network. However, in real-world networks, node attribute information is also available in addition to topology information. Considering more than one source of information for community detection could produce meaningful clusters and improve the robustness of the network. Therefore, a pre-processing approach that considers attribute information, shared neighbours and connectivity information aspects of the network for community detection is presented in this thesis as part of my research. Finally, a set of real-world mobile phone usage data obtained from Cambridge Laboratories (Device Analyzer) has been analysed as an exploratory step for viability to apply the algorithms developed in this thesis. All the proposed approaches have been evaluated and verified for feasibility using real-world large data set. The evaluation results of these experimentations prove very promising for the type of large data networks considered.

The proliferation of smart mobile devices and user applications has continued to contribute to the tremendous volume of data traffic in cellular networks. Moreover, with the feature of heterogeneous connectivity interfaces of these smart devices, it becomes more complex for managing the traffic volume in the context of mobility. To surmount this challenge, service and resource providers are looking for alternative mechanisms that can successfully facilitate managing network resources and mobility in a more dynamic, predictive and distributed manner. New concepts of network architectures such as Software-Defined Network (SDN) and Network Function Virtualization (NFV) have paved the way to move from static to flexible networks. They make networks more flexible (i.e., network providers capable of on-demand provisioning), easily customizable and cost effective. In this regard, network slicing is emerging as a new technology built on the concepts of SDN and NFV. It splits a network infrastructure into isolated virtual networks and allows them to manage network resources based on their requirements and characteristics. Most of the existing solutions for network slicing are facing challenges in terms of resource and mobility management. Regarding resource management, it creates challenges in terms of provisioning network throughput, end-to-end delay, and fairness resources allocation for each slice, whereas, in the case of mobility management, due to the rapid change of user mobility the network slice operator would like to hold the mobility controlling over its clients across different access networks, rather than the network operator, to ensure better services and user experience. In this thesis, we propose two novel architectural solutions to solve the challenges identified above. The first proposed solution introduces a Network Slicing Resource Management (NSRM) mechanism that assigns the required resources for each slice, taking into consideration resource isolation between different slices. The second proposed v solution provides a Mobility Management architecture-based Network Slicing (MMNS) where each slice manages its users across heterogeneous radio access technologies such as WiFi, LTE and 5G networks. In MMNS architecture, each slice has different mobility demands (e.g,. latency, speed and interference) and these demands are governed by a network slice configuration and service characteristics. In addition, NSRM ensures isolating, customizing and fair sharing of distributed bandwidths between various network slices and users belonging to the same slice depending on different requirements of each one. Whereas, MMNS is a logical platform that unifies different Radio Access Technologies (RATs) and allows all slices to share them in order to satisfy different slice mobility demands. We considered two software simulations, namely OPNET Modeler and OMNET++, to validate the performance evaluation of the thesis contributions. The simulation results for both proposed architectures show that, in case of NSRM, the resource blocking is approximately 35% less compared to the legacy LTE network, which it allows to accommodate more users. The NSRM also successfully maintains the isolation for both the inter and intra network slices. Moreover, the results show that the NSRM is able to run different scheduling mechanisms where each network slice guarantee perform its own scheduling mechanism and simultaneously with other slices. Regarding the MMNS, the results show the advantages of the proposed architecture that are the reduction of the tunnelling overhead and the minimization of the handover latency. The MMNS results show the packets delivery cost is optimal by reducing the number of hops that the packets transit between a source node and destination. Additionally, seamless session continues of a user IP-flow between different access networks interfaces has been successfully achieved.