Even though the e-learning is becoming increasingly popular in the academic environment, the quality of synchronous e-learning video is still substandard and significant work needs to be done to improve it. The improvements have to be brought about taking into considerations both: the network requirements and the psycho- physical aspects of the human visual system. One of the problems of the synchronous e-learning video is that the head-and-shoulder video of the instructor is mostly transmitted. This video presentation can be made more interesting by transmitting shots from different angles and zooms. Unfortunately, the transmission of such multi-shot videos will increase packet delay, jitter and other artifacts caused by frequent changes of the scenes. To some extent these problems may be reduced by controlled reduction of the quality of video so as to minimise uncontrolled corruption of the stream. Hence, there is a need for controlled streaming of a multi-shot e-learning video in response to the changing availability of the bandwidth, while utilising the available bandwidth to the maximum. The quality of transmitted video can be improved by removing the redundant background data and utilising the available bandwidth for sending high-resolution foreground information. While a number of schemes exist to identify and remove the background from the foreground, very few studies exist on the identification and separation of the two based on the understanding of the human visual system. Research has been carried out to define foreground and background in the context of e-learning video on the basis of human psychology. The results have been utilised to propose methods for improving the transmission of e-learning videos. In order to transmit the video sequence efficiently this research proposes the use of Feed- Forward Controllers that dynamically characterise the ongoing scene and adjust the streaming of video based on the availability of the bandwidth. In order to satisfy a number of receivers connected by varied bandwidth links in a heterogeneous environment, the use of Multi-Layer Feed-Forward Controller has been researched. This controller dynamically characterises the complexity (number of Macroblocks per frame) of the ongoing video sequence and combines it with the knowledge of availability of the bandwidth to various receivers to divide the video sequence into layers in an optimal way before transmitting it into network. The Single-layer Feed-Forward Controller inputs the complexity (Spatial Information and Temporal Information) of the on-going video sequence along with the availability of bandwidth to a receiver and adjusts the resolution and frame rate of individual scenes to transmit the sequence optimised to give the most acceptable perceptual quality within the bandwidth constraints. The performance of the Feed-Forward Controllers have been evaluated under simulated conditions and have been found to effectively regulate the streaming of real-time e-learning videos in order to provide perceptually improved video quality within the constraints of the available bandwidth.

C. Shannon presented theoretical conditions under which communication was possible error-free in the presence of noise. Subsequently the notion of using error correcting codes to mitigate the effects of noise in digital transmission was introduced by R. Hamming. Algebraic codes, codes described using powerful tools from algebra took to the fore early on in the search for good error correcting codes. Many classes of algebraic codes now exist and are known to have the best properties of any known classes of codes. An error correcting code can be described by three of its most important properties length, dimension and minimum distance. Given codes with the same length and dimension, one with the largest minimum distance will provide better error correction. As a result the research focuses on finding improved codes with better minimum distances than any known codes. Algebraic geometry codes are obtained from curves. They are a culmination of years of research into algebraic codes and generalise most known algebraic codes. Additionally they have exceptional distance properties as their lengths become arbitrarily large. Algebraic geometry codes are studied in great detail with special attention given to their construction and decoding. The practical performance of these codes is evaluated and compared with previously known codes in different communication channels. Furthermore many new codes that have better minimum distance to the best known codes with the same length and dimension are presented from a generalised construction of algebraic geometry codes. Goppa codes are also an important class of algebraic codes. A construction of binary extended Goppa codes is generalised to codes with nonbinary alphabets and as a result many new codes are found. This construction is shown as an efficient way to extend another well known class of algebraic codes, BCH codes. A generic method of shortening codes whilst increasing the minimum distance is generalised. An analysis of this method reveals a close relationship with methods of extending codes. Some new codes from Goppa codes are found by exploiting this relationship. Finally an extension method for BCH codes is presented and this method is shown be as good as a well known method of extension in certain cases.

The world has witnessed the development of many wireless systems and applications. In addition to the large number of existing devices, such development of new and advanced wireless systems increases rapidly the demand for more radio spectrum. The radio spectrum is a limited natural resource; however, it has been observed that it is not efficiently utilized. Consequently, different dynamic spectrum access techniques have been proposed as solutions for such an inefficient use of the spectrum. Cognitive Radio (CR) is a promising intelligent technology that can identify the unoccupied portions of spectrum and opportunistically uses those portions with satisfyingly high capacity and low interference to the primary users (i.e., licensed users). The CR can be distinguished from the classical radio systems mainly by its awareness about its surrounding radio frequency environment. The spectrum sensing task is the main key for such awareness. Due to many advantages, Orthogonal Frequency Division Multiplexing system (OFDM) has been proposed as a potential candidate for the CR‟s physical layer. Additionally, the Fast Fourier Transform (FFT) in an OFDM receiver supports the performance of a wide band spectrum analysis. Multitaper spectrum estimation method (MTM) is a non-coherent promising spectrum sensing technique. It tolerates problems related to bad biasing and large variance of power estimates. This thesis focuses, generally, on the local, multi antenna based, and global cooperative spectrum sensing techniques at physical layer in OFDM-based CR systems. It starts with an investigation on the performance of using MTM and MTM with singular value decomposition in CR networks using simulation. The Optimal MTM parameters are then found. The optimal MTM based detector theoretical formulae are derived. Different optimal and suboptimal multi antenna based spectrum sensing techniques are proposed to improve the local spectrum sensing performance. Finally, a new concept of cooperative spectrum sensing is introduced, and new strategies are proposed to optimize the hard cooperative spectrum sensing in CR networks. The MTM performance is controlled by the half time bandwidth product and number of tapers. In this thesis, such parameters have been optimized using Monte Carlo simulation. The binary hypothesis test, here, is developed to ensure that the effect of choosing optimum MTM parameters is based upon performance evaluation. The results show how these optimal parameters give the highest performance with minimum complexity when MTM is used locally at CR. The optimal MTM based detector has been derived using Neyman-Pearson criterion. That includes probabilities of detection, false alarm and misses detection approximate derivations in different wireless environments. The threshold and number of sensed samples controlling is based on this theoretical work. In order to improve the local spectrum sensing performance at each CR, in the CR network, multi antenna spectrum sensing techniques are proposed using MTM and MTM with singular value decomposition in this thesis. The statistical theoretical formulae of the proposed techniques are derived including the different probabilities. The proposed techniques include optimal, that requires prior information about the primary user signal, and two suboptimal multi antenna spectrum sensing techniques having similar performances with different computation complexity; these do not need prior information about the primary user signalling. The work here includes derivations for the periodogram multi antenna case. Finally, in hard cooperative spectrum sensing, the cooperation optimization is necessary to improve the overall performance, and/or minimize the number of data to be sent to the main CR-base station. In this thesis, a new optimization method based on optimizing the number of locally sensed samples at each CR is proposed with two different strategies. Furthermore, the different factors that affect the hard cooperative spectrum sensing optimization are investigated and analysed and a new cooperation scheme in spectrum sensing, the master node, is proposed.

Vehicle Ad hoc Networks (V ANET) have been proposed to provide safety on the road and deliver road traffic information and route guidance to drivers along with commercial applications. However the challenges facing V ANET are numerous. Nodes move at high speeds, road side units and basestations are scarce, the topology is constrained by the road geometry and changes rapidly, and the number of nodes peaks suddenly in traffic jams. In this thesis we investigate the physical and link layers of V ANET and propose methods to achieve high data rates and high throughput. For the physical layer, we examine the use of Vertical BLAST (VB LAST) systems as they provide higher capacities than single antenna systems in rich fading environments. To study the applicability of VB LAST to VANET, a channel model was developed and verified using measurement data available in the literature. For no to medium line of sight, VBLAST systems provide high data rates. However the performance drops as the line of sight strength increases due to the correlation between the antennas. Moreover, the performance of VBLAST with training based channel estimation drops as the speed increases since the channel response changes rapidly. To update the channel state information matrix at the receiver, a channel tracking algorithm for flat fading channels was developed. The algorithm updates the channel matrix thus reducing the mean square error of the estimation and improving the bit error rate (BER). The analysis of VBLAST-OFDM systems showed they experience an error floor due to inter-carrier interference (lCI) which increases with speed, number of antennas transmitting and number of subcarriers used. The update algorithm was extended to VBLAST -OFDM systems and it showed improvements in BER performance but still experienced an error floor. An algorithm to equalise the ICI contribution of adjacent subcarriers was then developed and evaluated. The ICI equalisation algorithm reduces the error floor in BER as more subcarriers are equalised at the expense of more hardware complexity. The connectivity of V ANET was investigated and it was found that for single lane roads, car densities of 7 cars per communication range are sufficient to achieve high connectivity within the city whereas 12 cars per communication range are required for highways. Multilane roads require higher densities since cars tend to cluster in groups. Junctions and turns have lower connectivity than straight roads due to disconnections at the turns. Although higher densities improve the connectivity and, hence, the performance of the network layer, it leads to poor performance at the link layer. The IEEE 802.11 p MAC layer standard under development for V ANET uses a variant of Carrier Sense Multiple Access (CSMA). 802.11 protocols were analysed mathematically and via simulations and the results prove the saturation throughput of the basic access method drops as the number of nodes increases thus yielding very low throughput in congested areas. RTS/CTS access provides higher throughput but it applies only to unicast transmissions. To overcome the limitations of 802.11 protocols, we designed a protocol known as SOFT MAC which combines Space, Orthogonal Frequency and Time multiple access techniques. In SOFT MAC the road is divided into cells and each cell is allocated a unique group of subcarriers. Within a cell, nodes share the available subcarriers using a combination of TDMA and CSMA. The throughput analysis of SOFT MAC showed it has superior throughput compared to the basic access and similar to the RTS/CTS access of 802.11.

We present a method to recover the reflectance of objects and the parameters of multiple lights illuminating the scene using a 3D image acquired by a depth sensor and a stereo intensity pair. The methodology has three components: the object geometric model, the surface reflection properties and the light source parameters. In the simpler case, we allow calibration of the point light sources using a Lambertian sphere; this can be used subsequently to estimate the surface reflection properties of an object with known geometry, acquired by an active sensor and represented as a triangulation mesh. However, the main contribution of the thesis is to remove the need for pre-calibration, and estimate the light source parameters, the diffuse and specular surface reflection parameters and the surface geometry from image data alone. By adapting a welldetermined iterative regression process on the illumination equation, a Gaussian Subtraction (GS) algorithm is used to separate the diffuse and specular components of the surface reflectance across the object surface. Then, the locations and intensities of the several light sources are recovered. We demonstrate the approach on both synthetic and real data acquired from a series of objects of reasonable complexity, using up to three light sources. No pre-calibration of the surfaces or of the light positions is necessary. For evaluation, we compare the results to ground truth when available, but also render these objects in dynamic X3DNRML format for a subjective comparison.

There is an increasing amount of information being made available as data streams, e.g. stock tickers, data from sensor networks, smart homes, monitoring data, etc. In .many cases, this data is generated by distributed sources under the control of many different organisations. Users would like to seamlessly query such data without prior knowledge of where it is located or how it is published..This is similar to the problem of integrating data residing in multiple heterogeneous stored data sources. However, the techniques developed for stored data are not applicable due to the continuous and long-lived nature of queries over data streams. This thesis proposes an architecture for a stream integration system. A key feature of the architect~e is a republisher component that collects together distributed streams and makes the merged ~tream available for querying. A formal model for the system has been developed. and is used to generate plans for executing continuous queries which exploit the redundancy introduced by the republishers. Additionally, due to the long-lived nature of continuous queries, mechanisms for maintaining the plans whenever there is a change in tlie set of data sources have been developed. A prototype of the system has been implemented and performance measures made. The work of this thesis has been motivated by the problem of retrieving monitoring information about Grid resources. However, the techniques developed are general and can be applied wherever there is a need to publish and query distributed data involving data streams.

No description available.

No description available.

No description available.

Networks of phased array radars are able to provide better counter stealth target detection and classification. Each radar sensor/node generates information which requires transmission to a central control authority who is able to evaluate the information. This requires a communications network to be established to allow transmission of information to and from any node. Each radar node is limited by range and degree and relies on the formation of a multi-hop network to facilitate this transmission. This thesis investigates a method whereby the radar beam itself is used in the formation of a multi-hop network. The phased array's multifunctional nature allows rapid switching between communications and radar function. A complete model of how the communication system could operate is included in this thesis. In order to simulate radar function and network communication, a custom built simulation platform was used. Two different approaches to the formation of the radar networks are derived, one involving a global design method using the Strength Pareto Evolutionary Algorithm (SPEA) and another using a local method specifically tailored to radar network formation, Distributed Algorithm for Radar Topology Control (DARTC). Both of these techniques can be used in the formation of radar networks. The thesis goes further to suggest modifications to the model itself that could result in improved radar network communication performance.