To improve the performance of adaptive algorithms, we develop algorithms adapted on the noise characteristics rather than adapting only on second order statistics. The developments in this thesis can be classified in two major works. First work is on developing a minimum bit-error rate (MBER) decision feedback equaliser (DFE) for impulsive noise modelled as a α-stable distribution. The development exploits the stable nature of the α-distribution and the concepts build on earlier work in a Gaussian noise environment. Further, a Wiender-filter-with-limiter solution is also presented and used as a performance bench mark. An improvement in convergence and BER performance is achieved by using a minimum bit error rate (MBER) cost function instead of a conventional least mean square (LMS) based design. The ability of least BER (LBER) equalisers based on a Gaussian noise assumption to operate in α-stable noise environment is also highlighted. In the second work, a block based maximum-likelihood algorithm using kernel density estimates to improve channel estimation in non-Gaussian noise environment is proposed. The likelihood pdf is assumed unknown and is estimated by using a kernel density estimator at the receiver. Thereby combining log-likelihood as a cost function with a kernel density estimator provides a robust channel estimator, which could be used for various non-Gaussian noise environments without any modification. The performance of the proposed estimator is compared with the theoretical lower bounds for associated noise distribution. The simulations for impulsive noise and co-channel interference (CCI) in presence of Gaussian noise, confirms that a better estimate can be obtained by using the proposed technique as compared to the traditional algorithms. The proposed algorithm is then applied to orthogonal frequency division multiplexing (OFDM) communication systems. A considerable performance improvement is observed when using a non-parametric channel estimator in conjunction with a symbol-by-symbol non-parametric maximum <i>a posteriori</i> probability (MAP) equaliser.

This thesis investigates the use of semi-continuous hidden Markov models (HMM) for automatic speaker verification (ASV) over a telephone channel. The system which was implemented is evaluated on a large database of isolated digits recorded over the British telephone network. The goal of the work is to improve performance of the ASV system under the constraints of limited enrolment data (5 tokens of each digit) and realistic computational and storage requirements. Experiments are conducted on the combined use of several standard feature sets under a common state segmentation, multiple codebook architecture. The feature sets investigated are linear predictive cepstral coefficients, mel-frequency cepstral coefficients and their respective first order differences. New algorithms which are proposed and evaluated include the weighting of digits scores according to their usefulness to the verification task and using Gaussian state duration probabilities as an additional information source in the verification decision. The most important contribution of this thesis is the development of a method for the construction of discriminating HMMs without the need for discriminative training. This new form of model, known as a discriminating observation probability (DOP) HMM involves the combination of standard HMMs to form a discriminating model. The DOP models are more flexible and perform better than the <I>speaker normalisation </I>techniques which are currently favoured in the literature. DOP models have potential application to many binary classification tasks using HMMs. The equal error rate (EER) using speaker specific thresholds on a series of 12 isolated digits was 0.17% using multiple codebook DOP models, compared to 1.93% using single codebook conventional HMM models. This represents a reduction in EER of 91%.

The thesis studies the following problem: given a set of geometrical figures (such as planar polygons), each one labelled according to whether or not it resembles some 'ideal' figure, find a good approximation to that ideal figure which can be used to classify other figures in the same way. We work within the PAC learning model introduced by Valiant in 1984. Informally, the concepts under consideration are sets of polygons which resemble each other visually. A learning algorithm is given collections of members and non-members of a concept, and its task is to infer a criterion for membership which is consistent with the given examples and which can be used as an accurate classifier of further example polygons. In order to formalise the notion of a concept, we use metrics which measure the extent to which two polygons differ. A concept is assumed to be the set of polygons which are within some distance of some fixed central polygon. In the thesis we work most extensively with the Hausdorff metric. Using the Hausdorff metric we obtain NP-completeness results for several variants of the learning problem. In particular we show that it is hard to find a single geometrical figure which is close to the positive examples but not to the negative examples. This result holds under various assumptions about the specific geometrical figures under consideration. It also holds for several metrics other than the Hausdorff metric. Despite the NP-completeness results mentioned above we have found some encouraging positive results. In particular, we have discovered a general technique for <i>prediction</i>. (Prediction is a less demanding learning model than PAC learning. The goal is to find a polynomial-time algorithm which takes as input a sample of labelled examples and is then able to predict the status of further unlabelled examples in polynomial time).

No description available.

No description available.

Cooperative communication becomes a practical alternative to multiple-input multiple-output (MIMO) systems when MIMO cannot be implemented due to size, cost or hardware limitations. The diversity provided by cooperative communication techniques can effectively combat the deleterious effect of fading, thus cooperative diversity is believed as a powerful approach to facilitate high-speed and high-quality wireless communications. Low-Density Parity-Check LDPC) codes were largely ignored until Mackay and Neal rediscovered in 1995. Nowadays owing to its sparse graph structure and low complexity iterative decoding methods, LDPC codes have not only been successful in the field of coding theory but are also being used in a variety of practical applications. This thesis explores practical decoding schemes of LDPC codes for cooperative communication. We start with the Gaussian multiple access channel (MAC) since the MAC could be seen as one component in a multi-user cooperative communication system. In this part, we investigate the performance of LDPC codes with soft iterative successive interference cancellation (SIC) and decoding scheme using BPSK modulation. Next, we propose a cooperative diversity scheme for the communication model of two sources sharing a single relay under two scenarios. The scheme uses algebraic code superposition relating in a frequency division mode or in the multiple access fading channel to create spatial diversity under the constraint of limited communication resources. We also describe in detail a novel, computationally efficient message passing algorithm at the destinations decoder which can extract the substantial spatial diversity contained in the code and signal superposition. The decoder is based on a sliding window structure where certain a posteriori soft values are retained to form a priori soft values for the next decoding. We show that despite the simplicity of the proposed scheme, diversity gains are efficiently leveraged by the simple combination of channel coding at the sources and network coding at the relay.

Energy efficient Media Access Control (MAC) strategies have been widely confirmed to be one of the essential solutions for power-saving in Mobile Ad Hoc Networks (MANETs). The aim of the work in this thesis is to examine the existing energy efficient MAC protocols for wireless ad hoc networks, and to develop enhanced energy efficient MAC strategies for MANETs. Taking a typical example of the WLAN ad hoc network power saving mode, a statistical model is proposed for the analysis of the power-saving MAC performance in the presence of Rayleigh fading. The analysis and simulation results show that the power-saving MAC mechanisms will degrade networking performance, such as end-to-end delay, and throughput, on the other hand, the impact of power-saving MAC methods on routing protocols is studied based on the proposed analytical model. Consequently, new power saving MAC mechanisms are proposed to balance the power saving and throughput under dynamic traffic load and improve the routing discovery performance. Power-saving MAC protocols are also strongly linked to the network topology management. Based on the comparison and discussion of various wireless network topologies, a novel hierarchical clustering algorithm is proposed in this thesis to prolong the lifetime of the whole network, improve the network throughput and simplify topology management in a large scale network. In MANETs, because of mobility, the dynamic radio environment, and the distributed networking operations, power control is different from those in other type of networks. Focusing on the solutions of power control in the MAC layer, this thesis also presents a novel power control and interference mitigation method for Wireless Personal Area Networks (WPANs). The method proposed is based on the Piconet Coordinator (PNC) selection mechanism in IEEE802.15-based WPANs, since the PNC is important as it centrally controls all networking operations. A practical implementation of the proposed power control method is also introduced in this thesis. The experiments demonstrate that the proposed method can dramatically improve the energy efficiency and network robustness.

This thesis reports on an investigation of various system architectures and receiver structures for cellular communications systems which discriminate users by direct sequence code division multiple access (DS-CDMA). Attention is focussed on the downlink of such a spread spectrum system and the influence of a number of design parameters is considered. The objective of the thesis is to investigate signal processing techniques which may be employed either at the receiver, or throughout the system to improve the overall capacity. The principles of spread spectrum communication are first outlined, including a discussion of the relative merits of spreading sequence sets, and a description of various signal processing techniques which are to be applied to the multi-user environment. The measure of system performance is introduced, and the conventional DS-CDMA system is analysed theoretically and through simulation to provide a reference performance level. Adaptive algorithms, which iteratively approximate the minimum mean square error (Wiener) receiver filter, are then investigated, both in stationary additive white Gaussian noise (AWGN) and in a more realistic radio channel. The inter-dependence of the system chip-rate, maximum Doppler offset induced by the motion of the receiver and the tuning parameters of the adaptive algorithm are demonstrated. Aspects of forward error correction (FEC) coding are then investigated, with convolutional coding on the data used both as an alternative to and as a supplement of direct sequence spreading. The most efficient use of the available expansion in bandwidth is shown to be dependent on a balance between FEC coding power and the capacity of the spreading sequence set chosen. Methods of combining multiple access interference cancellation techniques with convolutional coding and Viterbi decoding are considered. New structures are proposed which incorporate FEC decoding at the intermediate stage of the canceller, and the performance of these receivers is analysed theoretically. Simulations confirm that significant capacity improvements may be achieved, at tolerable increases in computational complexity and processing delay.

In the past, a number of researchers have applied various spectral estimation techniques in an attempt to analyse recorded heart sounds. The majority of these studies have used spectral estimation algorithms such as the Fourier transform and various autoregressive modelling techniques. Despite the definite potential these techniques have shown for the diagnosis of valvular heart disease, they are limited by their assumption of signal stationary and lack of relation to present stethoscope-based medical evaluation procedures. A solution to these limitations can be achieved by analysing the recorded sounds in the time-frequency domain rather than in the frequency-domain of time-domain independently. The research detailed in this thesis investigates the application of time-frequency techniques to the description and analysis of recorded heart sounds. Time-frequency is further investigated as a tool for the description of heart sounds in an attempt to diagnose valvular heart disease. Data used in the study was recorded from 100 subjects in four main valve populations. The four populations investigated were subjects with native heart valves, Carpentier-Edwards bioprosthetic heart valves, Bjork-Shiley metallic prosthetic heart valves and subjects before and after surgery for heart valve replacement. Prior to the analysis of these data sets, an investigation was performed into the suitability of various time-frequency techniques to the analysis of heart sounds. By comparing the short-time Fourier transform, wavelet transform, Wigner distribution and the Choi-Williams distribution it was found that the Choi-Williams distribution provides definite advantages over the other techniques due to its high resolution and reduced interference properties. Applying the Choi-Williams distribution to typical examples of each data set demonstrated that time-frequency offers definite potential as a heart sound descriptor. Typical results also demonstrate that time-frequency can be used as an aid to understanding the origins of heart sounds.

In this thesis we investigate MIMO techniques for relay channels. We first concentrate on the single antenna relay networks, where each node is equipped with a single antenna. In this scenario, a scheme called cooperative diversity has been discussed widely, where multiple relays are united in the network as a “virtual antenna array”, to mimic a MIMO system. We propose a novel cooperative diversity scheme that can improve the spectral efficiency of the network, especially for high signal to noise ratios (SNR). We analyze the capacity bounds for such schemes and also describe a signalling method to approach this capacity bound. <br> We then move to the multi-antenna node scenario, where each node is equipped with multiple antennas. We propose different signalling methods and routing protocols for MIMO relay channels and use capacity as a performance metric to evaluate and compare them. The proposed signalling methods can be applied together with the proposed routing schemes. Incorporating them can facilitate the cross-layer design. Finally, we discuss a network scenario where some nodes are equipped with multiple antennas, others are equipped with single antennas. We constrain ourselves to the case where the source and destination are equipped with a single antenna. We characterize the capacity performance and the diversity-multiplexing tradeoff of such a network. We show that relaying can offer a significant performance advantage over non-relay transmission in certain scenarios, by applying signal combining techniques for the point-to-point MIMO link into relay channels.