The increasing use of interactive multimedia applications over the Internet has created a problem of congestion. This is because a majority of these applications do not respond to congestion indicators. This leads to resource starvation for responsive flows, and ultimately excessive delay and losses for all flows therefore loss of quality. This results in unfair sharing of network resources and increasing the risk of network ‘congestion collapse’. Current Congestion Control Mechanisms such as ‘TCP-Friendly Rate Control’ (TFRC) have been able to achieve ‘fair-share’ of network resource when competing with responsive flows such as TCP, but TFRC’s method of congestion response (i.e. to reduce Packet Rate) is not ideally matched for interactive multimedia applications which maintain a fixed Frame Rate. This mismatch of the two rates (Packet Rate and Frame Rate) leads to buffering of frames at the Sender Buffer resulting in delay and loss, and an unacceptable reduction of quality or complete loss of service for the end-user. To address this issue, this thesis proposes a novel Congestion Control Mechanism which is referred to as ‘TCP-friendly rate control – Fine Grain Scalable’ (TFGS) for interactive multimedia applications. This new approach allows multimedia frames (data) to be sent as soon as they are generated, so that the multimedia frames can reach the destination as quickly as possible, in order to provide an isochronous interactive service. This is done by maintaining the Packet Rate of the Congestion Control Mechanism (CCM) at a level equivalent to the Frame Rate of the Multimedia Encoder.The response to congestion is to truncate the Packet Size, hence reducing the overall bitrate of the multimedia stream. This functionality of the Congestion Control Mechanism is referred to as Packet Size Truncation (PST), and takes advantage of adaptive multimedia encoding, such as Fine Grain Scalable (FGS), where the multimedia frame is encoded in order of significance, Most to Least Significant Bits. The Multimedia Adaptation Manager (MAM) truncates the multimedia frame to the size indicated by the Packet Size Truncation function of the CCM, accurately mapping user demand to available network resource. Additionally Fine Grain Scalable encoding can offer scalability at byte level granularity, providing a true match to available network resources. This approach has the benefits of achieving a ‘fair-share’ of network resource when competing with responsive flows (as similar to TFRC CCM), but it also provides an isochronous service which is of crucial benefit to real-time interactive services. Furthermore, results illustrate that an increased number of interactive multimedia flows (such as voice) can be carried over congested networks whilst maintaining a quality level equivalent to that of a standard landline telephone. This is because the loss and delay arising from the buffering of frames at the Sender Buffer is completely removed. Packets sent maintain a fixed inter-packet-gap-spacing (IPGS). This results in a majority of packets arriving at the receiving end at tight time intervals. Hence, this avoids the need of using large Playout (de-jitter) Buffer sizes and adaptive Playout Buffer configurations. As a result this reduces delay, improves interactivity and Quality of Experience (QoE) of the multimedia application.

The continuous miniaturisation of sensors, as well as the progression in wearable electronics, embedded software, digital signal processing and biomedical technologies, have led to new usercentric networks, where devices can be carried in the user’s pockets, attached to the user’s body. Body-centric wireless communications (BCWCs) is a central point in the development of fourth generation mobile communications. Body-centric wireless networks take their place within the personal area networks, body area networks and sensor networks which are all emerging technologies that have a wide range of applications (such as, healthcare, entertainment, surveillance, emergency, sports and military). The major difference between BCWC and conventional wireless systems is the radio channels over which the communication takes place. The human body is a hostile environment from a radio propagation perspective and it is therefore important to understand and characterise the effects of the human body on the antenna elements, the radio channel parameters and, hence, system performance. This thesis focuses on the study of body-worn antennas and on-body radio propagation channels. The performance parameters of five different narrowband (2.45 GHz) and four UWB (3.1- 10.6 GHz) body-worn antennas in the presence of human body are investigated and compared. This was performed through a combination of numerical simulations and measurement campaigns. Parametric studies and statistical analysis, addressing the human body effects on the performance parameters of different types of narrowband and UWB antennas have been presented. The aim of this study is to understand the human body effects on the antenna parameters and specify the suitable antenna in BCWCs at both 2.45 GHz and UWB frequencies. Extensive experimental investigations are carried out to study the effects of various antenna types on the on-body radio propagation channels as well. Results and analysis emphasize the best body-worn antenna for reliable and power-efficient on-body communications. Based on the results and analysis, a novel dual-band and dual-mode antenna is proposed for power-efficient and reliable on-body and off-body communications. The on-body performance of the DBDM antenna at 2.45 GHz is compared with other five narrowband antennas. Based on the results and analysis of six narrowband and four UWB antennas, antenna specifications and design guidelines are provided that will help in selecting the best body-worn antenna for both narrowband and UWB systems to be applied in body-centric wireless networks (BCWNs). A comparison between IV the narrowband and UWB antenna parameters are also provided. At the end of the thesis, the subject-specificity of the on-body radio propagation channel at 2.45 GHz and 3-10 GHz was experimentally investigated by considering eight real human test subjects of different shapes, heights and sizes. The subject-specificity of the on-body radio propagation channels was compared between the narrowband and UWB systems as well.

Optimizing radio resource within a network and across cooperating heterogeneous networks is the focus of this thesis. Cooperation in a multi-network environment is tackled by investigating network selection mechanisms. These play an important role in ensuring quality of service for users in a multi-network environment. Churning of mobile users from one service provider to another is already common when people change contracts and in a heterogeneous communication environment, where mobile users have freedom to choose the best wireless service-real time selection is expected to become common feature. This real time selection impacts both the technical and the economic aspects of wireless network operations. Next generation wireless networks will enable a dynamic environment whereby the nodes of the same or even different network operator can interact and cooperate to improve their performance. Cooperation has emerged as a novel communication paradigm that can yield tremendous performance gains from the physical layer all the way up to the application layer. Game theory and in particular coalitional game theory is a highly suited mathematical tool for modelling cooperation between wireless networks and is investigated in this thesis. In this thesis, the churning behaviour of wireless service users is modelled by using evolutionary game theory in the context of WLAN access points and WiMAX networks. This approach illustrates how to improve the user perceived QoS in heterogeneous networks using a two-layered optimization. The top layer views the problem of prediction of the network that would be chosen by a user where the criteria are offered bit rate, price, mobility support and reputation. At the second level, conditional on the strategies chosen by the users, the network provider hypothetically, reconfigures the network, subject to the network constraints of bandwidth and acceptable SNR and optimizes the network coverage to support users who would otherwise not be serviced adequately. This forms an iterative cycle until a solution that optimizes the user satisfaction subject to the adjustments that the network provider can make to mitigate the binding constraints, is found and applied to the real network. The evolutionary equilibrium, which is used to 3 compute the average number of users choosing each wireless service, is taken as the solution. This thesis also proposes a fair and practical cooperation framework in which the base stations belonging to the same network provider cooperate, to serve each other‘s customers. How this cooperation can potentially increase their aggregate payoffs through efficient utilization of resources is shown for the case of dynamic frequency allocation. This cooperation framework needs to intelligently determine the cooperating partner and provide a rational basis for sharing aggregate payoff between the cooperative partners for the stability of the coalition. The optimum cooperation strategy, which involves the allocations of the channels to mobile customers, can be obtained as solutions of linear programming optimizations.

Mobile terminals allow users to access services while on the move. This unique feature has driven the rapid growth in the mobile network industry, changing it from a new technology into a massive industry within less than two decades. Handover is the essential functionality for dealing with the mobility of the mobile users. Compared with the conventional hard handover employed in the GSM mobile networks, the soft handover used in IS-95 and being proposed for 3G has better performance on both link and system level. Previous work on soft handover has led to several algorithms being proposed and extensive research has been conducted on the performance analysis and parameters optimisation of these algorithms. Most of the previous analysis focused on the uplink direction. However, in future mobile networks, the downlink is more likely to be the bottleneck of the system capacity because of the asymmetric nature of new services, such as Internet traffic. In this thesis, an in-depth study of the soft handover effects on the downlink direction of WCDMA networks is carried out, leading to a new method of optimising soft handover for maximising the downlink capacity and a new power control approach.

This thesis investigates reinforcement learning algorithms suitable for learning in large state space problems and coevolution. In order to learn in large state spaces, the state space must be collapsed to a computationally feasible size and then generalised about. This thesis presents two new implementations of the classic temporal difference (TD) reinforcement learning algorithm Sarsa that utilise fuzzy logic principles for approximation, FQ Sarsa and Fuzzy Sarsa. The effectiveness of these two fuzzy reinforcement learning algorithms is investigated in the context of an agent marketplace. It presents a practical investigation into the design of fuzzy membership functions and tile coding schemas. A critical analysis of the fuzzy algorithms to a related technique in function approximation, a coarse coding approach called tile coding is given in the context of three different simulation environments; the mountain-car problem, a predator/prey gridworld and an agent marketplace. A further comparison between Fuzzy Sarsa and tile coding in the context of the nonstationary environments of the agent marketplace and predator/prey gridworld is presented. This thesis shows that the Fuzzy Sarsa algorithm achieves a significant reduction of state space over traditional Sarsa, without loss of the finer detail that the FQ Sarsa algorithm experiences. It also shows that Fuzzy Sarsa and gradient descent Sarsa(λ) with tile coding learn similar levels of distinction against a stationary strategy. Finally, this thesis demonstrates that Fuzzy Sarsa performs better in a competitive multiagent domain than the tile coding solution.

A Mobile Ad Hoc Networks (MANET) is a collection of mobile nodes that can communicate with each other using multihop wireless links without utilizing any fixed based-station infrastructure and centralized management. Each mobile node in the network acts as both a host generating flows or being destination of flows and a router forwarding flows directed to other nodes. Future applications of MANETs are expected to be based on all-IP architecture and be capable of carrying multitude real-time multimedia applications such as voice and video as well as data. It is very necessary for MANETs to have an efficient routing and quality of service (QoS) mechanism to support diverse applications. This thesis proposes an on-demand Node-Disjoint Multipath Routing protocol (NDMR) with low broadcast redundancy. Multipath routing allows the establishment of multiple paths between a single source and single destination node. It is also beneficial to avoid traffic congestion and frequent link breaks in communication because of the mobility of nodes. The important components of the protocol, such as path accumulation, decreasing routing overhead and selecting node-disjoint paths, are explained. Because the new protocol significantly reduces the total number of Route Request packets, this results in an increased delivery ratio, smaller end-to-end delays for data packets, lower control overhead and fewer collisions of packets. Although NDMR provides node-disjoint multipath routing with low route overhead in MANETs, it is only a best-effort routing approach, which is not enough to support QoS. DiffServ is a standard approach for a more scalable way to achieve QoS in any IP network and could potentially be used to provide QoS in MANETs because it minimises the need for signalling. However, one of the biggest drawbacks of DiffServ is that the QoS provisioning is separate from the routing process. This thesis presents a Multipath QoS Routing protocol for 111 supporting DiffServ (MQRD), which combines the advantages of NDMR and DiffServ. The protocol can classify network traffic into different priority levels and apply priority scheduling and queuing management mechanisms to obtain QoS guarantees.

Object tracking is one of the fundamental tasks in various applications such as surveillance, sports, video conferencing and activity recognition. Factors such as occlusions, illumination changes and limited field of observance of the sensor make tracking a challenging task. To overcome these challenges the focus of this thesis is on using multiple modalities such as audio and video for multi-target, multi-modal tracking. Particularly, this thesis presents contributions to four related research topics, namely, pre-processing of input signals to reduce noise, multi-modal tracking, simultaneous detection and tracking, and interaction recognition. To improve the performance of detection algorithms, especially in the presence of noise, this thesis investigate filtering of the input data through spatio-temporal feature analysis as well as through frequency band analysis. The pre-processed data from multiple modalities is then fused within Particle filtering (PF). To further minimise the discrepancy between the real and the estimated positions, we propose a strategy that associates the hypotheses and the measurements with a real target, using a Weighted Probabilistic Data Association (WPDA). Since the filtering involved in the detection process reduces the available information and is inapplicable on low signal-to-noise ratio data, we investigate simultaneous detection and tracking approaches and propose a multi-target track-beforedetect Particle filtering (MT-TBD-PF). The proposed MT-TBD-PF algorithm bypasses the detection step and performs tracking in the raw signal. Finally, we apply the proposed multi-modal tracking to recognise interactions between targets in regions within, as well as outside the cameras’ fields of view. The efficiency of the proposed approaches are demonstrated on large uni-modal, multi-modal and multi-sensor scenarios from real world detections, tracking and event recognition datasets and through participation in evaluation campaigns.

A mobile ad hoc network (MANET) is a collection of mobile nodes that can communicate with each other without using any fixed infrastructure. It is necessary for MANETs to have efficient routing protocol and quality of service (QoS) mechanism to support multimedia applications such as video and voice. Node-Disjoint Multipath Routing Protocol (NDMR) is a practical protocol in MANETs: it reduces routing overhead dramatically and achieves multiple node-disjoint routing paths. Because QoS support in MANETs is important as best-effort routing is not efficient for supporting multimedia applications, this thesis presents a novel approach to provide that support. In this thesis NDMR is enhanced to provide a QoS enabled NDMR that decreases the transmission delay between source and destination nodes. A multi-rate mechanism is also implemented in the new protocol so that the NDMR QoS can minimise the overall delays. It is shown that these approaches lead to significant performance gains. A modification to NDMR is also proposed to overcome some of the limitations of the original.

It is widely expected that next-generation wireless communication systems will be heterogeneous, integrating a wide variety of wireless access networks. Of particular interest recently is a mix of cellular networks (GSM/GPRS and WCDMA) and wireless local area networks (WLANs) to provide complementary features in terms of coverage, capacity and mobility support. If cellular/ WLAN interworking is to be the basis for a heterogeneous network then the analysis of complex handover traffic rates in the system (especially vertical handover) is one of the most essential issues to be considered. This thesis describes the application of queueing-network theory to the modelling of this heterogeneous wireless overlay system. A network of queues (or queueing network) is a powerful mathematical tool in the performance evaluation of many large-scale engineering systems. It has been used in the modelling of hierarchically structured cellular wireless networks with much success, including queueing network modelling in the study of cellular/ WLAN interworking systems. In the process of queueing network modelling, obtaining the network topology of a system is usually the first step in the construction of a good model, but this topology analysis has never before been used in the handover traffic study in heterogeneous overlay wireless networks. In this thesis, a new topology scheme to facilitate the analysis of handover traffic is proposed. The structural similarity between hierarchical cellular structure and heterogeneous wireless overlay networks is also compared. By replacing the microcells with WLANs in a hierarchical structure, the interworking system is modelled as an open network of Erlang loss systems and with the new topology, the performance measures of blocking probabilities and dropping probabilities can be determined. Both homogeneous and non-homogeneous traffic have been considered, circuit switched and packet-switched. Example scenarios have been used to validate the models, the numerical results showing clear agreement with the known validation scenarios.

This thesis explores how audio can be better incorporated into how people access information and does so by developing approaches for creating three-dimensional audio environments with low processing demands. This is done by investigating three research questions. Mobile applications have processor and memory requirements that restrict the number of concurrent static or moving sound sources that can be rendered with binaural audio. Is there a more e cient approach that is as perceptually accurate as the traditional method? This thesis concludes that virtual Ambisonics is an ef cient and accurate means to render a binaural auditory display consisting of noise signals placed on the horizontal plane without head tracking. Virtual Ambisonics is then more e cient than convolution of HRTFs if more than two sound sources are concurrently rendered or if movement of the sources or head tracking is implemented. Complex acoustics models require signi cant amounts of memory and processing. If the memory and processor loads for a model are too large for a particular device, that model cannot be interactive in real-time. What steps can be taken to allow a complex room model to be interactive by using less memory and decreasing the computational load? This thesis presents a new reverberation model based on hybrid reverberation which uses a collection of B-format IRs. A new metric for determining the mixing time of a room is developed and interpolation between early re ections is investigated. Though hybrid reverberation typically uses a recursive lter such as a FDN for the late reverberation, an average late reverberation tail is instead synthesised for convolution reverberation. Commercial interfaces for music search and discovery use little aural information even though the information being sought is audio. How can audio be used in interfaces for music search and discovery? This thesis looks at 20 interfaces and determines that several themes emerge from past interfaces. These include using a two or three-dimensional space to explore a music collection, allowing concurrent playback of multiple sources, and tools such as auras to control how much information is presented. A new interface, the amblr, is developed because virtual two-dimensional spaces populated by music have been a common approach, but not yet a perfected one. The amblr is also interpreted as an art installation which was visited by approximately 1000 people over 5 days. The installation maps the virtual space created by the amblr to a physical space.