The success of many image restoration algorithms is often due to their ability to sparsely describe the original signal. Many sparse promoting transforms exist, including wavelets, the so called ‘lets’ family of transforms and more recent non-local learned transforms. The first part of this thesis reviews sparse approximation theory, particularly in relation to 2-D piecewise polynomial signals. We also show the connection between this theory and current state of the art algorithms that cover the following image restoration and enhancement applications: denoising, deconvolution, interpolation and multi-view super resolution. In [63], Shukla et al. proposed a compression algorithm, based on a sparse quadtree decomposition model, which could optimally represent piecewise polynomial images. In the second part of this thesis we adapt this model to image restoration by changing the rate-distortion penalty to a description-length penalty. Moreover, one of the major drawbacks of this type of approximation is the computational complexity required to find a suitable subspace for each node of the quadtree. We address this issue by searching for a suitable subspace much more efficiently using the mathematics of updating matrix factorisations. Novel algorithms are developed to tackle the four problems previously mentioned. Simulation results indicate that we beat state of the art results when the original signal is in the model (e.g. depth images) and are competitive for natural images when the degradation is high.

The renewable policies in various countries are driving significant growth of grid connected renewable generation sources such as the Photovoltaics (PVs). Typically a PV generation is integrated into power systems at the low and the medium voltage distribution level. The uptake of an intermittent power from the PVs is challenging the power system operation and control. The network voltage control is one of the major challenges during the operation of the distribution connected PVs. The active power injection from a PV plant causes variable voltage rise. This forces the existing voltage control devices such as on-load tap-changer (OLTC) and voltage regulator (VR) to operate continuously. The consequence is the reduction of the operating life of the voltage control mechanism. Also, the conventional non-coordinated reactive power control results in the operation of the VR at its control limit (VR runaway condition). This research focuses on the distribution voltage control in the presence of PV generation and helps to establish detailed insights into the various associated challenges. Firstly, the typical grid integrated PV topologies are discussed. The existing power system operational practices are presented and their limitations are identified. A voltage control methodology to tackle challenges such as over-voltage, excessive tap counts and VR runaway is presented. These challenges are alleviated through the coordinated reactive power control. The reactive power coordination is achieved through the deterministic distribution optimal power flow solved through the interior point technique. The irradiance and the load forecasting errors are another set of challenges from the distribution network operators' perspective. The stochastic optimal voltage control strategy is proposed to tackle the element of randomness associated with the forecast errors. The stochastic operational risks such as an over- voltage and a VR runaway are defined through a chance constrained optimization problem. The simulation study is performed using a realistic 95-bus UK generic distribution network model and a practically measured irradiance to demonstrate the effectiveness of the proposed control strategies. The thesis makes an effort to offer an insight into the operational challenges and propose strategies to achieve a seamless integration of the PVs into the power systems.

People are the centre of technologies. Understanding, monitoring and tracking the behaviour of people will benefit in various areas including driving assistance, surveillance for safety and caring purposes and applications for machine-people interaction. Particularly, pedestrians attract more attention for two reasons: they restrict the behaviours of people to standing and moving upright; and applications for pedestrian detection and monitoring have positively impact on the quality of life. Pedestrian detection and identification, aims at recognising pedestrians fromstill images and video frames. Together with pedestrian recognition and tracking, this topic attempts to train computers to recognise a pedestrian. The problem is challenging. Though frameworks were designed, various algorithms were proposed in recent years, further efforts are needed to improve the accuracy and reliability of the performance. In this thesis, proposing a modifiable framework for pedestrian identification and improving the performances of current pedestrian detection techniques are particularly focused. Based on appearance based pedestrian identification, a modifiable framework is a novel philosophy of developing frameworks which can be easily improved. For pedestrian identification, a novel protocol where layers of algorithms are hierarchically applied to solve the problem. To compare the detected pedestrians, appearance based features are selected, the "bag-of-features" framework is employed to compare the histogram descriptions of pedestrians. To improve the performances of HOG pedestrian detector, the presence of head-shoulder structure is selected as the evidence of the presence of pedestrian. A novel appearance based framework is developed to detect the head-shoulder structure from the detection results of HOG detector. Furthermore, in order to separate multiple pedestrians detected in one bounding box, a novel algorithm is proposed to detect the approximated symmetry axes of pedestrians.

There are two main challenges when producing in vitro cell systems: first, to reconstitute the in situ cellular microenvironment, thus delivering more representative and reliable cell models for drug screening and disease modelling studies. Second, to record and quantify the electrical and chemical gradients across the culture. Ideally, both challenges should be accomplished within a single platform towards a lab-on-chip implementation. This research work investigates the application of Parylene C in cardiac cell scaffolding and its integrability with electrochemical monitoring technologies for measuring extracellular action potentials and pH. The surface properties of Parylene C in terms of water affinity, chemical composition and nanotopography were characterised before and after modifying the material's inherent hydrophobicity through oxygen plasma. A technology was developed to selectively alter the surface hydrophobicity of Parylene C through standard lithography and oxygen plasma, which is characterised by μm-resolution and long-term pattern stability, and can accurately control the extent of induced hydrophilicity, the pattern layout and 3-D geometry. The micro-engineered Parylene C films were employed as scaffolds for cardiac cells with immature physiological properties, such as neonatal rat ventricular myocytes (NRVM). The scaffolds promoted a more in situ cellular structure and organisation, while they improved important calcium (Ca2+) cycling parameters such as fluorescent amplitude, time to peak (Tp), time to 50% (T50) and 90% (T90) decay at 0.5-2 Hz field stimulation. The thickness of the patterned Parylene C films was found to regulate the shape of the cells by controlling their adhesion area on the Parylene substrate through a thickness-dependent hydrophobicity. NRVM on thin (2 μm) membranes tended to bridge across the hydrophobic areas and adopt a spread-out shape (average contact angle at the level of the nucleus was 64.51o). On the other hand, cells on thick (10 μm) films were mostly constrained on the hydrophilic areas and demonstrated a more elongated, cylindrical (in vivo-like) shape (average contact angle was 84.73o). The cylindrical shape and a significantly (p <0.05) denser microtubule structure in cells on thick films possibly suggest a more mature cardiomyocyte. However, there was no significant effect on the Ca2+ physiology between the two groups. The micro-patterning technology was able to deliver free-standing Parylene C thin films (2-10 μm) to study the effect of substrate elasticity and flexibility on the Ca2+ physiology of NRVM. Preliminary results showed that fluorescent amplitude and time to peak were improved in structured NRVM cultures on stand-alone Parylene films compared to rigid Parylene-coated glass surfaces. However, no such trend was present in Ca2+ release parameters (T50, T90). The flexibility of the culture substrate was also manipulated by employing free-standing micro-patterned Parylene C films of distinct thicknesses (2-10 μm), but did not affect the cellular Ca2+ physiology. Further biological validation is needed with a larger sample size to draw a certain conclusion. The cell patterning technology was transferred to commercially available planar Multi-Electrode arrays (MEAs) to demonstrate integrability of this method with existing monitoring tools. The micro-patterned MEAs induced anisotropic cardiomyocyte cultures, as they substantially increased the longitudinal-to-transverse velocity anisotropy ratios (1.09, n=4 to 1.69, n=2), promoting action potential propagation profiles that closer resembled native cardiac tissue. Furthermore, the micro-engineered MEAs were proven to be reusable, yielding a versatile and low-cost approach that is compatible with state-of-art recording equipment and can be employed as a more reliable, off-the-shelf tool for drug screening studies. Selective hydrophilic modification of Parylene C was also employed to activate locally the H+ sensing capacity of such films, implementing extended-gate pH sensors. The ability of Parylene C to act in a dual way - as an encapsulation material and as an active pH sensing membrane - was demonstrated. The material exhibited a distinguishable sensitivity dependent on the oxygen plasma recipe, relatively low drift rates and excellent encapsulation quality. Based on these principles, flexible Parylene-based high-density miniaturised electrode arrays were fabricated, employing Parylene as a flexible structure material and as a H+ sensing membrane for local detection of pH. The presented Parylene-based technology has the potential to deliver integrated lab-on-chip implementations for growing cells in vitro with controlled microtopography while monitoring the extracellular electrical and pH gradients across the culture in a non-invasive way, with application in drug screening and disease modelling.

Increased penetration of distributed generations in distribution networks are altering the technical characteristics of the grid, pushing them to operate closer to their limits of safe and reliable operation. New renewable generators connecting to the distribution network will be constrained due to the presence of thermal and voltage constraints during times of low demand and high generation output. The traditional reinforcement planning by means of increasing the capacity of network assets can be very costly and usually ends up in overinvested network with low utilization rates of the assets. In recent years, some smart technologies have been introduced to be used to increase the utilization of network assets and provide the adequate capacity for Distributed Generations (DGs). These smart solutions can help the Distributed Network Operators (DNOs) to provide cheaper and faster network connections for DGs. This thesis presents a multi epoch Optimal Power Flow (OPF) model for capacity and voltage management of a distribution network for integrating new DGs. The model uses the smart solutions including Dynamic Line Rating (DLR), Quad-Booster (QB), Static VAR Compensator (SVC) and Automatic Network Management (ANM) for DGs as well as the traditional reinforcement options. Also the model finds the optimal connection points for new DGs to reduce the cost of network investment and DG curtailment. The multi epoch model is solved with both incremental approach where the investment is carried out incrementally and with integrated approach where the planning is done strategically anticipating the future needs of the network. It compares the application of smart solutions in short and long term planning. The proposed model is applied to a generic UK distribution network and the results are discussed.

Objective: Diagnosis of sleep disorders is an expensive procedure that requires performing a sleep study, known as polysomnography (PSG), in a controlled environment. This study monitors the neural, eye and muscle activity of a patient using electroencephalogram (EEG), electrooculogram (EOG) and electromyogram (EMG) signals which are then scored in to different sleep stages. Home PSG is often cited as an alternative of clinical PSG to make it more accessible, however it still requires patients to use a cumbersome system with multiple recording channels that need to be precisely placed. This thesis proposes a wearable sleep staging system using a single channel of EEG. For realisation of such a system, this thesis presents novel features for REM sleep detection from EEG (normally detected using EMG/EOG), a low-complexity automatic sleep staging algorithm using a single EEG channel and its complete integrated circuit implementation. Methods: The difference between Spectral Edge Frequencies (SEF) at 95% and 50% in the 8-16 Hz frequency band is shown to have high discriminatory ability for detecting REM sleep stages. This feature, together with other spectral features from single-channel EEG are used with a set of decision trees controlled by a state machine for classification. The hardware for the complete algorithm is designed using low-power techniques and implemented on chip using 0.18μm process node technology. Results: The use of SEF features from one channel of EEG resulted in 83% of REM sleep epochs being correctly detected. The automatic sleep staging algorithm, based on contextually aware decision trees, resulted in an accuracy of up to 79% on a large dataset. Its hardware implementation, which is also the very first complete circuit level implementation of any sleep staging algorithm, resulted in an accuracy of 98.7% with great potential for use in fully wearable sleep systems.

The design of the compute elements of hardware, its datapath, plays a crucial role in determining the speed, area and power consumption of a device. The building blocks of datapath are polynomial in nature. Research into the implementation of adders and multipliers has a long history and developments in this area will continue. Despite such efficient building block implementations, correctly determining the necessary precision of each building block within a design is a challenge. It is typical that standard or uniform precisions are chosen, such as the IEEE floating point precisions. The hardware quality of the datapath is inextricably linked to the precisions of which it is composed. There is, however, another essential element that determines hardware quality, namely that of the accuracy of the components. If one were to implement each of the official IEEE rounding modes, significant differences in hardware quality would be found. But in the same fashion that standard precisions may be unnecessarily chosen, it is typical that components may be constructed to return one of these correctly rounded results, where in fact such accuracy is far from necessary. Unfortunately if a lesser accuracy is permissible then the techniques that exist to reduce hardware implementation cost by exploiting such freedom invariably produce an error with extremely difficult to determine properties. This thesis addresses the problem of how to construct hardware to efficiently implement fixed and floating-point polynomials while exploiting a global error freedom. This is a form of lossy synthesis. The fixed-point contributions include resource minimisation when implementing mutually exclusive polynomials, the construction of minimal lossy components with guaranteed worst case error and a technique for efficient composition of such components. Contributions are also made to how a floating-point polynomial can be implemented with guaranteed relative error.

Intraspinal microstimulation (ISMS) is an emerging method that is applied to neuroprosthesis aimed at individuals with spinal cord injury. Compared to traditional spinal stimulation or peripheral nerve stimulation methods, ISMS can activate muscle groups in organised synergies and thus can provide finer control of the generated force with reduced muscle fatigue. As the spinal cord is the neural link between the central and peripheral nervous systems, it is convenient to use this in accessing neurons associated with limb movement within a small area. For example, the relevant length of the spinal cord controlling the lower limbs in humans is only 5 cm. However, this means that any implant surgery is limited to some extent, on the other hand, this means ISMS needs to use invasive electric neural stimulation (ENS) with microelectrodes to access the target motor neurons to achieve a higher spatial resolution. Similar to other implantable ENS systems, an ISMS system needs to be compact, safe and energy efficient (in addition to effectively provide the required therapy). Although existing implantable neural stimulators fulfil these basic requirements, there is still much room for improvement. Depending on whether the stimulus is current or voltage controlled, the stimulator can be good for either safety and controllability or energy efficiency. Since the trend in the semiconductor industry is to reduce power consumption in integrated circuits, a current controlled stimulator is usually preferable by experimental neuroscientists. However, there is a new trend to combine these two control modalities, to enjoy the benefits of both. Following this trend, this thesis starts by focusing on a third modality -- charge controlled stimulation, which delivers the stimulus in charge packets. This eliminates the voltage headroom required for relatively high output resistances in current controlled stimulators whilst preserving the controllability over the total charge delivered. Charge controlled stimulation is thus proposed for having the potential to be as energy efficient as voltage controlled stimulation and as safe as current controlled stimulator. A novel circuit for charge mode stimulation is described based on a charge metering approach that has been adopted from nuclear engineering. Experimental results demonstrate the feasibility of this approach and also identify the key challenges. This is then extended to a novel reconfigurable, multi-modal and multichannel stimulator circuit. This is the first integrated system to implement current, voltage and charge control stimulation within a reconfigurable channel architecture. This has been developed to investigate the effect of dynamic multipolar electrode reconfiguration with the aim of focusing or steering the stimulus. To this end, different stimulus delivery methods can be tested for multipolar spatial control. The concept of multipolar stimulation is then investigated from a theoretical standpoint. The ability to apply this in improving the spatial resolution in ISMS can be achieved by confining the stimulus spread (thus reducing destructive crosstalk). This method can also be used to shift the stimulus voltage field away from the delivering electrode so as to correct implant placement error during surgery. A theoretical computational model is developed to investigate the effect of dynamic multipolar electrode reconfiguration with the aim of focusing or steering the stimulus. It is intended, together with the developed multichannel stimulator, that this will be used in future to develop advanced multipolar strategies that can achieve spatial hyperacuity for ISMS, and more generally in ENS.

In data mining problems the representation or description of data plays a fundamental role, since it defines the set of essential properties for the extraction and characterisation of patterns. However, for the case of temporal data, such as time series and data streams, one outstanding issue when developing mining algorithms is finding an appropriate data description or representation. In this thesis two novel domain-independent representation frameworks for temporal data suitable for off-line and online mining tasks are formulated. First, a domain-independent temporal data representation framework based on a novel data description strategy which combines structural and statistical pattern recognition approaches is developed. The key idea here is to move the structural pattern recognition problem to the probability domain. This framework is composed of three general tasks: a) decomposing input temporal patterns into subpatterns in time or any other transformed domain (for instance, wavelet domain); b) mapping these subpatterns into the probability domain to find attributes of elemental probability subpatterns called primitives; and c) mining input temporal patterns according to the attributes of their corresponding probability domain subpatterns. This framework is referred to as Structural Generative Descriptions (SGDs). Two off-line and two online algorithmic instantiations of the proposed SGDs framework are then formulated: i) For the off-line case, the first instantiation is based on the use of Discrete Wavelet Transform (DWT) and Wavelet Density Estimators (WDE), while the second algorithm includes DWT and Finite Gaussian Mixtures. ii) For the online case, the first instantiation relies on an online implementation of DWT and a recursive version of WDE (RWDE), whereas the second algorithm is based on a multi-resolution exponentially weighted moving average filter and RWDE. The empirical evaluation of proposed SGDs-based algorithms is performed in the context of time series classification, for off-line algorithms, and in the context of change detection and clustering, for online algorithms. For this purpose, synthetic and publicly available real-world data are used. Additionally, a novel framework for multidimensional data stream evolution diagnosis incorporating RWDE into the context of Velocity Density Estimation (VDE) is formulated. Changes in streaming data and changes in their correlation structure are characterised by means of local and global evolution coefficients as well as by means of recursive correlation coefficients. The proposed VDE framework is evaluated using temperature data from the UK and air pollution data from Hong Kong.

This thesis presents a decentralized alternative to the centralized state-estimation and control technologies used in current power systems. Power systems span over vast geographical areas, and therefore require a robust and reliable communication network for centralized estimation and control. The supervisory control and data acquisition (SCADA) systems provide such a communication architecture and are currently employed for centralized estimation and control of power systems in a static manner. The SCADA systems operate at update rates which are not fast enough to provide appropriate estimation or control of transient or dynamic events occurring in power systems. Packet-switching based networked control system (NCS) is a faster alternative to SCADA systems, but it suffers from some other problems such as packet dropouts, random time delays and packet disordering. A stability analysis framework for NCS in power systems has been presented in the thesis considering these problems. Some other practical limitations and problems associated with real-time centralized estimation and control are computational bottlenecks, cyber threats and issues in acquiring system-wide parameters and measurements. The aforementioned problems can be solved by a decentralized methodology which only requires local parameters and measurements for estimation and control of a local unit in the system. The cumulative effect of control at all the units should be such that the global oscillations and instabilities in the power system are controlled. Such a decentralized methodology has been presented in the thesis. The method for decentralization is based on a new concept of `pseudo-inputs' in which some of measurements are treated as inputs. Unscented Kalman filtering (UKF) is applied on the decentralized system for dynamic state estimation (DSE). An extended linear quadratic regulator (ELQR) has been proposed for the optimal control of each local unit such that the whole power system is stabilized and all the oscillations are adequately damped. ELQR requires DSE as a prerequisite. The applicability of integrated system for dynamic estimation and control has been demonstrated on a model 16-machine 68-bus benchmark system.