No description available.

The possibility of using heavy rare earth metals and some of their alloys as pole piece materials for nanoscale electromagnets as in magnetic recording devices (hard drive disks) at cryogenic temperatures was evaluated. To set the frame an extensive literature review on high-moment materials and concepts is presented. Thin films of Permendur (Fe6%Co35 alloys), gadolinium (Gd), dysprosium (Dy), Dy50Gd50 binary alloys, terbium (Tb), holmium (Ho), DyRh binary alloys, and DyRhX (X = Fe, Ni, Co, Gd) ternary alloys were magnetron-sputtered and characterised for their chemical and crystal structure. In-plane magnetic properties were investigated over a temperature range of T = 3 -300 K by SQUID magnetometry, with an emphasis on the saturation magnetisation, i.e. the magnetic moment per unit volume, and points of magnetic transition such as Curie temperature. All heavy rare earth metals (Gd, Dy, DyGd, Tb, Ho) were found to be suitable write-pole candidates, for their saturation magnetisation exceeds that of Fe65Co35 in the cryogenic temperature regime. In terms of reasonably high magnetisation at moderate temperatures, DyGd and Tb are deemed the best materials. However, other write-pole relevant magnetic properties such as softness and initial permeability were unfavourable for all heavy rare earth metals. The DyRh binary and DyRhX ternary alloy system were found to be completely unsuitable, since they had very low saturation magnetisation, due to antiparallel coupling of the rare earth and the transition metals, as well as very low Curie points, i.e. operating temperatures of T -0 K. All thin films had saturation magnetisations slightly reduced, and transition temperatures slightly lower, than that of their bulk counterparts as a result of finite size effects.

This dissertation reports on experiments in the design of a free-ranging robotic vehicle with locomotor and spatial skills. These two skills are suggested to be fundamental to any general mobile agent. No attempt has been made to incorporate manipulation, and motivational issues are not treated (the system's objectives are specified by a human operator). Two schools of thought with regard to robot design prevail: they are the representation-based or <i>analytic</i>, and representation-free or <i>synthetic</i> schools. These guiding philosophies are compared and contrasted in terms of their approaches to locomotion and localisation. In the course of this work, it has been found that the smoothness of a mobile robot's locomotion benefits from adjusting its speed to the density of obstacles it is encountering. A reactive, synthetic-type procedure for achieving this is offered, based on the frequency at which obstacle detection interrupts the robot's navigational system. However, it will be shown that a predictive, representation-based approach is more efficient. On the other hand, it will emerge that a very limited representation of the robot's operating environment and sensing systems is sufficient to allow the robot to localise accurately anywhere within its operating space, based only on the goodness of match between mapped and currently sensed data. The scheme treats the infrared, sonic, and ultrasonic sensors of an autonomous mobile robot as devices which partition its operating space qualitatively. It will be shown that the infrared and ultrasonic systems combined lead to position estimates within 10 cm of where the robot actually is, on average. These systems are found to be superior to the sonic system, surprisingly, because they have distinct regions within which they do not work: it is better to have accurate sensors with dead zones than inaccurate ones which can be used anywhere. Since it is not obvious which aspects of robot behaviour are well-suited, and which are unsuited for representation versus representation-free implementation, a conclusion drawn from the work is that elements of both approaches bear further consideration, and might be usefully combined.

Speech signals acquired, transmitted or stored in non-ideal conditions are often degraded by one or more effects including, for example, additive noise. These degradations alter the signal properties in a manner that deteriorates the intelligibility or quality of the speech signal. In the law enforcement context such degradations are commonplace due to the limitations in the audio collection methodology, which is often required to be covert. In severe degradation conditions, the acquired signal may become unintelligible, losing its value in an investigation and in less severe conditions, a loss in signal quality may be encountered, which can lead to higher transcription time and cost. This thesis proposes a non-intrusive speech assessment framework from which algorithms for speech quality and intelligibility assessment are derived, to guide the collection and transcription of law enforcement audio. These methods are trained on a large database labelled using intrusive techniques (whose performance is verified with subjective scores) and shown to perform favorably when compared with existing non-intrusive techniques. Additionally, a non-intrusive CODEC identification and verification algorithm is developed which can identify a CODEC with an accuracy of 96.8 % and detect the presence of a CODEC with an accuracy higher than 97 % in the presence of additive noise. Finally, the speech description taxonomy framework is developed, with the aim of characterizing various aspects of a degraded speech signal, including the mechanism that results in a signal with particular characteristics, the vocabulary that can be used to describe those degradations and the measurable signal properties that can characterize the degradations. The taxonomy is implemented as a relational database that facilitates the modeling of the relationships between various attributes of a signal and promises to be a useful tool for training and guiding audio analysts.

The Bearings-only tracking problem is to estimate the state of a moving object from noisy observations of its direction relative to a sensor. The Kalman filter, which provides least squares estimates for linear Gaussian filtering problems is not directly applicable because of the highly nonlinear measurement function of the state, representing the bearings measurements and so other types of filters must be considered. The shifted Rayleigh filter (SRF) is a highly effective moment-matching bearings-only tracking algorithm which has been shown, in 2D, to achieve the accuracy of computationally demanding particle filters in situations where the well-known extended Kalman filter and unscented Kalman filter often fail. This thesis has two principal aims. The first is to develop accurate and computationally efficient algorithms for bearings-only tracking in 3D space. We propose algorithms based on the SRF, that allow tracking, in the presence of clutter, of both nonmaneuvering and maneuvering targets. Their performances are assessed, in relation to competing methods, in highly challenging tracking scenarios, where they are shown to match the accuracy of high-order sophisticated particle filters, at a fraction of the computational cost. The second is to design accurate and consistent algorithms for bearings-only simultaneous localization and mapping (SLAM). The difficulty of this problem, originating from the uncertainty in the position and orientation of the sensor, and the absence of range information of observed landmarks, motivates the use of advanced bearings-only tracking algorithms. We propose the quadrature-SRF SLAM algorithm, which is a moment-matching filter based on the SRF, that numerically evaluates the exact mean and covariance of the posterior. Simulations illustrate the accuracy and consistency of its estimates in a situation where a widely used moment-matching algorithm fails to produce consistent estimates. We also propose a Rao-Blackwellized SRF implementation of a particle filter, which, however, does not exhibit favorable consistency properties.

The research presented in this thesis is concerned with multi-target localisation in MIMO radar. In particular, the aim is to develop novel algorithms which can improve the performance of target localisation. Firstly, a general spatiotemporal received signal model for MIMO radar is formulated. When the targets' relative delays are negligible, the general model turns into the spatial only signal model in which, in order to enjoy the enhanced parameter identifiability brought by the waveform diversity, a combined approach based on the virtual array structure is proposed for the multiple targets' directions and path gains estimation. The virtual array structure enables the proposed approach to identify more targets with accurate estimation. Besides, inspired by STAR manifold in communications, a novel spatiotemporal signal model for MIMO radar is proposed, which enables the existing multi-target localisation methods designed for the spatial only model working for the spatiotemporal one. Secondly, the multi-target localisation of MIMO radar operating in an envi- ronment with closely located targets is concerned. In such a scenario, the mu- tual interferences among targets severely degrade the performance of the current multi-target parameter estimators. Thus, an optimisation which takes account of the suppression of the mutual interferences for multi-target parameter estimation is formulated and the solutions to it are derived. Thirdly, based on these solu- tions, two novel multi-target parameter estimators are presented. By suppressing the interferences in the estimation, both the proposed methods outperform the existing ones. Finally, for the purpose of exploiting the high directional gain provided by the Tx beamforming in the multi-target localisation of MIMO radar, a joint Tx and Rx multi-target localisation approach is proposed. The cooperation between the Tx beamforming and target localisation enables the proposed approach to achieve better performance for the localisation of multiple targets.

Typical design flows for digital hardware take, as their input, an abstract description of computation and data transfer between logical memories. No existing commercial high-level synthesis tool demonstrates the ability to map logical memory inferred from a high level language to external memory resources. This thesis develops techniques for doing this, specifically targeting off-chip dynamic memory (DRAM) devices. These are a commodity technology in widespread use with standardised interfaces. In use, the bandwidth of an external memory interface and the latency of memory requests asserted on it may become the bottleneck limiting the performance of a hardware design. Careful consideration of this is especially important when designing with DRAMs, whose latency and bandwidth characteristics depend upon the sequence of memory requests issued by a controller. Throughout the work presented here, we pursue exact compile-time methods for designing application-specific memory systems with a focus on guaranteeing predictable performance through static analysis. This contrasts with much of the surveyed existing work, which considers general purpose memory controllers and optimized policies which improve performance in experiments run using simulation of suites of benchmark codes. The work targets loop-nests within imperative source code, extracting a mathematical representation of the loop-nest statements and their associated memory accesses, referred to as the ‘Polytope Model’. We extend this mathematical representation to represent the physical DRAM ‘row’ and ‘column’ structures accessed when performing memory transfers. From this augmented representation, we can automatically derive DRAM controllers which buffer data in on-chip memory and transfer data in an efficient order. Buffering data and exploiting ‘reuse’ of data is shown to enable up to 50× reduction in the quantity of data transferred to external memory. The reordering of memory transactions exploiting knowledge of the physical layout of the DRAM device allowing to 4× improvement in the efficiency of those data transfers.

Forecasting financial time series is regarded as one of the most challenging applications of time series prediction due to their dynamic nature. However, it is the fundamental element of most investment activities thus attracting the attention of practitioners and researchers for many decades. The purpose of this research is to investigate and develop novel methods for the prediction of financial time series considering their dynamic nature. The predictive performance of asset prices time series themselves is exploited by applying digital signal processing methods to their historical observations. The novelty of the research lies in the design of predictive filters by maximising their spectrum flatness of forecast errors. The filters are then applied to forecast linear combinations of daily open, high, low and close prices of financial time series. Given the assumption that there are no structural breaks or switching regimes in a time series, the sufficient and necessary conditions that a time series can be predicted with zero errors by linear filters are examined. It is concluded that a band-limited time series can be predicted with zero errors by a predictive filter that has a constant magnitude response and constant group delay over the bandwidth of the time series. Because real world time series are not band-limited thus cannot be forecasted without errors, statistical tests of spectrum flatness which evaluate the departure of the spectral density from a constant value are introduced as measures of the predictability of time series. Properties of a time series are then investigated in the frequency domain using its spectrum flatness. A predictive filter is designed by maximising the error spectrum flatness that is equivalent to maximise the “whiteness” of forecast errors in the frequency domain. The focus is then placed on forecasting real world financial time series. By applying spectrum flatness tests, it is found that the property of the spectrum of a linear combination of daily open, high, low and close prices, which is called target prices, is different from that of a random walk process as there are much more low frequency components than high frequency ones in its spectrum. Therefore, an objective function is proposed to derive the target price time series from the historical observations of daily open, high, low and close prices. A predictive filter is then applied to obtain the one-step ahead forecast of the target prices, while profitable trading strategies are designed based on the forecast of target prices series. As a result, more than 70% success ratio could be achieved in terms of one-step ahead out-of-sample forecast of direction changes of the target price time series by taking the S&P500 index for example.

The problem of estimating sparse signals based on incomplete set of noiseless or noisy measurements has been investigated for a long time from different perspec- tives. In this dissertation, after the review of the theory of compressed sensing (CS) and existing structured sensing matrices, a new class of convolutional sensing matri- ces based on deterministic sequences are developed in the first part. The proposed matrices can achieve a near optimal bound with O(K log(N)) measurements for non-uniform recovery. Not only are they able to approximate compressible signals in the time domain, but they can also recover sparse signals in the frequency and discrete cosine transform domain. The candidates of the deterministic sequences include maximum length sequence (or called m-sequence), Golay's complementary sequence and Legendre sequence etc., which will be investigated respectively. In the second part, Golay-paired Hadamard matrices are introduced as structured sensing matrices, which are constructed from the Hadamard matrix, followed by diagonal Golay sequences. The properties and performances are analyzed in the following. Their strong structures ensure special isometry properties, and make them be easier applicable to hardware potentially. Finally, we exploit novel CS principles successfully in a few real applications, including radar imaging and dis- tributed source coding. The performance and the effectiveness of each scenario are verified in both theory and simulations.

Brand loyalty is tightly connected with competition for market share between brands, as it describes consumers' willingness to maintain compatibility between subsequent purchases. Imposed by exogenous or endogenous factors, the reasons for brand loyalty can be summarized in the switching cost. However, consumers do not live in isolation, instead they influence and get influenced by peers. Especially if influence propagates via "word-of-mouth" and not direct marketing, peer pressure becomes invaluable as consumers reach for compatibility with other consumers. The evolution of market share competition can be described by coordination games played in networks over multiple periods. Consumers, acting as the players of the coordination game, contemplate the quality of the available choices while being susceptible at others' influence. By considering switching costs, inertia in their actions is also introduced. An important issue therefore arises - if the available choices are qualitatively similar, how do the switching costs in combination with network effects affect the competition outcome? In this thesis, we address this issue by introducing switching costs into the standard coordination game; the switching cost game thus describes potential losses inflicted on players by changing - switching - strategies. We particularly concentrate our efforts on the proportions of strategies at equilibrium. Compared to the cost-free game, the switching costs are responsible for the emergence of a coexistence region where all available strategies coexist in equilibrium. Numerical results on various network structures verified the coexistence outcome, along with the effect of cost values and their symmetry on the determination of the coexistence limits. Next, we investigate the monotonicity of the competition evolution. We exploit the monotonous behaviour of the switching cost game limiting cases over time to draw an analogy with bootstrap percolation and benefit from existing analytical methods to calculate their final outcome and create bounds of the general case. We also show how these bounds depend on the switching costs. Finally, we examine the effects of switching costs on the critical behaviour of the system by studying the properties of stable clusters.