Many studies have shown that brain-computer interface (BCI) technology is a potentially powerful tool for the rehabilitation of various psychological and neurological conditions, including restoration of movement and treatment of neuropathic pain (NP). However, most of these studies rely on expensive equipment, limiting its application to labs and hospital environments. Therefore, making BCI applications more readily available to patients is the main focus of this thesis. The aim of this study is to develop and assess two inexpensive, wearable neurorehabilitation systems that can be used for patient managed home-based therapy and are based on a portable brain-computer interface (PBCI) for neurofeedback (NF) applications. Both systems are inspired by neurorehabilitation protocols that have been previously tested on patients using laboratory BCI technology. The brain-computer interface systems are based on a wireless EEG system called EPOC, a Windows PC tablet and custom-made software developed under Visual C++. Both of these systems consist of portable BCI, one for neurofeedback (BCI-NF) and the other for controlling functional electrical stimulation (BCI-FES). System development followed the standard steps of user-centred design, while system testing followed the procedures for adopting new services or technologies, aiming to increase the usability of the BCI system in a patient population. The assessment phase, and in particular the assessment of PBCI-NF, included a systematic analysis of the main requirements and barriers for providing home-based BCI as a patient service, including training and support. The results of these chapters provide important feedback on usage patterns and technical problems, which could not be collected based on patients’ BCI experiences in laboratory or clinical trials. The ability to self-regulate brain waves was tested on able-bodied participants and patients with NP. Within the user-centred design frame, the effectiveness, efficiency, and user acceptance of BCI-NF were demonstrated on patients. The treatment was found to be comparable with the effectiveness of widely used pain drugs, with 53% of patients experiencing a clinically significant reduction in pain. The feasibility BCI-FES study on able-bodied participants and SCI tetraplegic patients demonstrated a high success rate in recognising motor intention within a single training session. This demonstrates the intuitiveness of the BCI-FES protocol, making it potentially suitable for extended, patient-managed hand therapy. In conclusion, this thesis demonstrated that SCI patients are able to use a BCI system on their own or through help from their caregiver in a home environment. It also demonstrated that the NF treatment has a positive effect on the reduction of CNP on SCI patients. In addition, this thesis presents promising results of home-based BCI systems in the rehabilitation domain and presents the first step in developing and testing consumer-grade BCI systems for rehabilitation purposes.

This thesis focuses on developing new automatic techniques addressing three typical problems in digital histopathology image analysis, histochemical stain separation at pixel-level, cell classifications at region level, and histochemical score assessment at image level, with the aim of providing useful tools to help histopathologists in their decision making. First, we study a pixel-level problem, separating positive chemical stains. To realise the full potential of digital pathology, accurate and robust computer techniques for automatically detecting biomarkers play an important role. Traditional methods transform the colour histopathology images into a gray scale image and apply a single threshold to separate positively stained tissues from the background. In this thesis, we show that the colour distribution of the positive immunohistochemical stains varies with the level of luminance and that a single threshold will be impossible to separate positively stained tissues from other tissues, regardless how the colour pixels are transformed. Based on this observation, two novel luminance adaptive biomarker detection methods are proposed. The first, termed Luminance Adaptive Multi-Thresholding (LAMT) first separate the pixels according to their luminance levels and for each luminance level a separate threshold is found for detecting the positive stains. The second, termed Luminance Adaptive Random Forest (LARF) applies one of the most powerful machine learning models, random forest, as a base classifier to build an ensemble classifier for biomarker detection. Second, we study a cell-level problem, the cell classification task in pathology images. Two different classification models are proposed. The first model for HEp-2 cell pattern classification comes with a novel object-graph based feature, which decompose the binary image into primitive objects and represent them with a set of morphological feature. Work on cell classification is further extended using deep learning model termed Deep Autoencoding-Classification Network (DACN). The DACN model consists of an autoencoder and a conventional classification convolutional neural network (CNN) with the two sharing the same encoding pipeline. The DACN model is jointly optimized for the classification error and the image reconstruction error based on a multi-task learning procedure. We will present experiment results to show that the proposed DACN outperforms all known state-of-the-art on two public indirect immunofluorescence stained HEp-2 cell datasets and H\&E stained colorectal adenocarcinomas cell dataset. Third, we study an image-level problem, assessing the histochemical score of a histopathology image. To determine the molecular class of the tumour, pathologists will have to manually mark the nuclei activity biomarkers by assigning a histochemical score (H-Score) to each TMA core with a semi-quantitative assessment method. Manually marking positively stained nuclei is a time consuming, imprecise and subjective process which will lead to inter-observer and intra-observer discrepancies. In this thesis, we present an end-to-end deep learning system which directly predicts the H-Score automatically. Our system imitates the pathologists' decision process and uses one fully convolutional network (FCN) to extract all nuclei region, a second FCN to extract tumour nuclei region, and a multi-column convolutional neural network which takes the outputs of the first two FCNs and the stain intensity description image as input and acts as the decision making mechanism to directly output the H-Score of the input TMA image. To the best of our knowledge, this is the first end-to-end system that takes a TMA image as input and directly outputs a clinical score. We will present experimental results which demonstrate that the H-Scores predicted by our model have very high and statistically significant correlation with experienced pathologists' scores and that the H-Score discrepancy between our algorithm and the pathologists is on par with the inter-subject discrepancy between the pathologists.

Collaborative learning is known as an effective learning method and various different kinds of technologies have been developed to support and facilitate collaborative learning. Many of these technologies are used to support the functional activities of a group of learners by enabling students to communicate, share documents and materials, track the work of the group, or distribute and allocate tasks. One factor that influences the success of collaborative groups is the awareness that members have of each others' activities i.e. activity awareness (Gutwin et al., 2004). Limited attention has been paid to promoting activity awareness in the collaborative learning literature. The work that does exist has focused on enhancing activity awareness by capturing and sharing details of the activity (e.g. Ganoe et al., 2003; Carroll et al., 2003). In contrast, there are no technologies that focus on the learners’ attitudes and behaviours with regard to activity awareness without considering the functional aspects of the group's work. This PhD hypothesises that persuasive technologies can offer a novel way of promoting activity awareness by changing learners’ attitudes and behaviours and persuading them to be more aware of fellow group members’ activities. This approach to enhancing activity awareness was investigated by using a persuasive social actor to change the attitudes and behaviours of learners who were working on collaborative learning projects over extended periods of time. Four studies were conducted: a pilot study to explore collaborative learning groups, an exploratory study to understand collaboration and activity awareness, a follow-up study to study activity awareness in depth, and a main study where a persuasive social actor for activity awareness in collaborative learning groups was developed and tested. All of these studies focused on a specific collaborative learning setting, in which small numbers of students (3 to 5) worked together in collaborative groups to complete real learning projects over approximately 6 weeks. This thesis makes four contributions to the fields of HCI and collaborative learning. The main contribution is a novel approach to enhance activity awareness in collaborative learning groups by changing learners’ attitudes and behaviours using a persuasive technology i.e. a persuasive social actor. The second contribution is a new method to evaluate activity awareness in collaborative learning groups. The third contribution is insight into how the Persuasive Systems Design (PSD) model (Oinas-kukkonen & Harjumaa, 2009) can be used in the design and evaluation of a persuasive social actor. The fourth contribution is an analysis of how students collaborate in long-term collaborative learning projects in naturalistic settings.

Computational models of classical conditioning have made significant contributions to the theoretic understanding of associative learning, yet they still struggle when the temporal aspects of conditioning are taken into account. Interval timing models have contributed a rich variety of time representations and provided accurate predictions for the timing of responses, but they usually have little to say about associative learning. In this thesis we present a unified model of conditioning and timing that is based on the influential Rescorla-Wagner conditioning model and the more recently developed Timing Drift-Diffusion model. We test the model by simulating 11 experimental phenomena and show that it can provide an adequate account for 9, and a partial account for the other 2. We argue that the model can account for more phenomena in the chosen set than these other similar in scope models: CSCTD, MS-TD, Learning to Time and Modular Theory. A comparison and analysis of the mechanisms in these models is provided, with a focus on the types of time representation and associative learning rule used.

Mobile user interfaces are becoming increasingly complex due to the expanding range of functionalities that they incorporate, which poses significant difficulties in software development. Formal methods are beneficial for highly complex software systems, as they enable the designed behaviour of a mobile user interface (UI) to be modelled and tested for accuracy before implementation. Indeed, assessing the compatibility between the software specification and user requirements and verifying the implementation in relation to the specification are essential procedures in the development process of any type of UI. To ensure that UIs meet users‘ requirements and competences, approaches that are based on interaction between humans and computers employ a variety of methods to address key issues. The development of underlying system functionality and UIs benefit from formal methods as well as from user-interface design specifications. Therefore, both approaches are incorporated into the software development process in this thesis. However, this integration is not an easy task due to the discrepancies between the two approaches. It also includes a method, which can be applied for both simple and complex UI applications. To overcome the issue of integrating both approaches, the thesis proposes a new formal model called the Formal Model of Mobile User Interface Design (FMMUID). This model is devised to characterise the composition of the UI design based on hierarchical structure and a set theory language. To determine its applicability and validity, the FMMUID is implemented in two real-world case studies: the quiz game iPlayCode and the social media application Social Communication (SC). A comparative analysis is undertaken between two case studies, where each case study has three existing applications with similar functionality in terms of structure and numbers of elements, functions and colours. Furthermore, the case studies are also assessed from a human viewpoint, which reveals that they possess better usability. The assessment supports the viability of the proposed model as a guiding tool for software development. The efficiency of the proposed model is confirmed by the result that the two case studies are less complex than the other UI applications in terms of hierarchical structure and numbers of elements, functions and colours, whilst also presenting acceptable usability in terms of the four examined dimensions: usefulness, information quality, interface quality, and overall satisfaction. Hence, the proposed model can facilitate the development process of mobile UI applications.

Centrifugal pumps play an important role in engineering applications since they are commonly used in industrial and residential systems, covering wide range of flow rates. Improving the performance of turbomachines such as the centrifugal pumps can be difficult to achieve, since the flow is turbulent with unsteady behaviour and cavitation. Cavitation is a complex phenomenon that is commonly considered as one of the main causes of deterioration in pump performance. Diagnosing cavitation and detecting its level of severity are essential for maintaining the pump’s reliability. Continuous condition monitoring of the pump is important to increase its operational life, decrease maintenance costs and hence, enhance the reliability of the pump. Early detection of cavitation can also improve the pump’s life expectancy by adopting various preventative actions. In this research, the first technique used for detecting cavitation is Computational Fluid Dynamics because it can provide suitable visualisation and reasonably accurate information, regarding the behaviour of fluid flow in the pump. In this work, both qualitative and quantitative analyses were carried out through a wide range of operating conditions and different geometrical configurations of a centrifugal pump under single-phase and cavitation conditions. Both, global and local flow field characteristics were investigated for better understanding. For qualitatively analysis, contours of static pressure and velocity magnitude under single-phase conditions and vapour volume fractions contour under cavitation conditions were adopted. On the other hand, the head and pressure variation in both time and frequency domains were analysed for qualitative analysis. The results showed that, as the pump rotational speed, number of impeller blades, and the outlet impeller diameter increase the head of the pump increases as well as the occurrence of cavitation. Based on the extensive numerical investigations for variety of operational and geometrical parameters, novel semi-empirical correlations under single-phase and cavitation conditions for the pump head and power coefficients were developed. Developments of aforementioned relations were carried out using multiple regression analysis technique. The second and third research areas consist of an extensive experimental analysis on the effects of operating conditions on the pump performance to predict cavitation using vibration and acoustic signature analyses. Detailed experimental investigations were carried out for the detection and diagnosis of cavitation, with the aid of sophisticated equipment and sensors. The condition monitoring was experimentally carried out in both, time and frequency domains analyses. Time domain method was applied to analyse the vibration and acoustic signals in time waveform analysis (TWFA). These signatures were analysed using different statistical parameters such as peak, root mean square (RMS), peak-to-peak and variance. In addition, transforming and analysing these signals in frequency domain was made by using Fast Fourier Transform technique. Analyses of these signals in frequency domain were also carried out using different statistical parameters such as mean and RMS features under wide various frequency ranges. The results revealed that using cavitation detection index (CDI) was a powerful technique, which can be used in both time and frequency domains for detecting cavitation and comparing the sensitivity of the vibration and acoustic techniques in estimating earlier stage of cavitation. Moreover, vibration technique was more sensitive to detect different levels of cavitation, especially inception of cavitation as compared to acoustic technique. This research has also found that the range of frequency between 0Hz to 15kHz was more sensitive to detect cavitation in the pump at the early stages. However, further investigation indicated that a frequency range of 1Hz to 2kHz was also effective on predicting the cavitation. Based on these findings, it can be suggested to use low range of frequency sensors (accelerometer and microphone) to capture the cavitation phenomenon instead of higher range of frequency, which are more expensive. In addition, it was found that all three techniques adopted in this investigation such as; CFD, vibration and acoustic techniques are well capable to analyse cavitation behaviours under different operating conditions. Moreover, it was observed that the numerical results and vibration technique can detect the inception of cavitation within a pump earlier than the acoustic technique. The results also revealed that, the combined use of these techniques (numerical and experimental) could increase the reliability. The combined method can be a considered to be a robust method, which can provide detailed information about the performance of the pump and detection/diagnosis of cavitation within a centrifugal pump. Hence, this will assist in prolonging the life of the pump and protect the system from emergency shutdown.

Condition monitoring (CM) of gear transmission has attracted extensive research in recent years. In particular, the detection and diagnosis of its faults in their early stages to minimise cost by maximising time available for planned maintenance and giving greater opportunity for avoiding a system breakdown. However, the diagnostic results obtained from monitored signals are often unsatisfactory because mainstream technologies using vibration response do not sufficiently account for the effect of friction and lubrication. To develop a more advanced and accurate diagnosis, this research has focused on investigating the nonlinearities of vibration generation and transmission with the viscoelastic properties of lubrication, to provide an in-depth understanding of vibration generating mechanisms and hence develop more effective signal processing methods for early detection and accurate diagnosis of gear incipient faults. A comprehensive dynamic model has been developed to study the dynamic responses of a multistage helical gear transmission system. It includes not only time-varying stiffness but also tooth friction forces based on an elastohydrodynamic lubrication (EHL) model. In addition, the progression of a light wear process is modelled by reducing stiffness function profile, in which the 2nd and 3rd harmonics of the meshing frequency (and their sidebands) show significant alteration that support fault diagnostic at early stages. Numerical and experimental results show that the friction and progressive wear induced vibration excitations will change slightly the amplitudes of the spectral peaks at both the mesh frequency and its sideband components at different orders, which provides theoretical supports for extracting reliable diagnostic signatures. As such changes in vibrations are extremely small and submerged in noise, it is clear that effective techniques for enhancing the signal-to-noise ratio, such as time synchronous averaging (TSA) and modulation signal bispectrum (MSB) are required to reveal such changes. MSB is preferred as it allows small amplitude sidebands to be accurately characterised in a nonlinear way without information loss and does not impose any addition demands regarding angular displacement measurement as does TSA. With the successful diagnosis of slight wear in helical gears, the research progressed to validate the capability of MSB based methods to diagnose four common gear faults relating to gear tribological conditions; lubrication shortfall, changes in lubrication viscosity, water in oil, and increased bearing clearances. The results show that MSB signatures allows accurate differentiation between these small changes, confirming the model and signal processing proposed in this thesis.

Extensible Markup Language (XML) has become a dominant technology for transferring data through the worldwide web. The XML labelling schemes play a key role in handling XML data efficiently and robustly. Thus, many labelling schemes have been proposed. However, these labelling schemes have limitations and shortcomings. Thus, the aim of this research was to investigate the existing XML labelling schemes and their limitations in order to address the issue of efficiency of XML query performance. This thesis investigated the existing labelling schemes and classified them into three categories based on certain criteria, in order to identify the limitations and challenges of these labelling schemes. Based on the outcomes of this investigation, this thesis proposed a state-of-theart labelling scheme, called clustering-based labelling scheme, to resolve or improve the key limitations such as the efficiency of the XML query processing, labelling XML nodes, and XML updates cost. This thesis argued that using certain existing labelling schemes to label nodes, and using the clustering-based techniques can improve query and labelling nodes efficiency. Theoretically, the proposed scheme is based on dividing the nodes of an XML document into clusters. Two existing labelling schemes, which are the Dewey and LLS labelling schemes, were selected for labelling these clusters and their nodes. Subsequently, the proposed scheme was designed and implemented. In addition, the Dewey and LLS labelling scheme were implemented for the purpose of evaluating the proposed scheme. Subsequently, four experiments were designed in order to test the proposed scheme against the Dewey and LLS labelling schemes. The results of these experiments suggest that the proposed scheme achieved better results than the Dewey and LLS schemes. Consequently, the research hypothesis was accepted overall with few exceptions, and the proposed scheme showed an improvement in the performance and all the targeted features and aspects.

Measurement of the velocity profile of water continuous multiphase flows is important because it can enable production optimisation and avoidance of unwanted flow assurance issues in both mining and oil and gas industries. However, accurate measurement of the velocity profile of the continuous phase in multiphase flows when they are time dependent or transient is still a challenge in such industries. Many available commercial multiphase flow meters, which are not able to directly measure the velocity profile of the continuous phase, use radioactive measurement techniques. Radioactive measurement techniques have many safety issues involving exposure to radiation, which is very harmful and is a known cause of cancer in humans. This thesis describes the development of a non-radioactive based flow meter, which relies on the measurement principle of a multi-electrode electromagnetic flow meter. This flow meter is capable of measuring the velocity profile of the conducting continuous phase in both single and multiphase flows tens (or potentially even hundreds) of times every second. The images of velocity profile of the conducting continuous phase in both single and multiphase flows were reconstructed using inductive flow tomography technique (IFT) from flow induced potential difference measurements obtained from a flush mounted array of electrodes on the wall of the flow meter. The designed and developed IFT system presented in this thesis consists of (i) a flow meter body, which has coils for generating magnetic fields and an array of 16-electrodes to enable sensing of flow induced potential differences; (ii) analogue electronic circuits for coil excitation and for signal conditioning of flow induced potential difference measurements and (iii) a computer unit for controlling system hardware and data acquisition and processing (which includes the mathematical algorithm for reconstructing the velocity profile of the continuous phase). Performance of the IFT system was tested in vertical single-phase ‘water only’ flows and in both vertical and inclined two-phase air-in-water and solids-in-water flows. The velocity profile measurements from the IFT system were in good agreement with reference measurements and were consistent with previous work cited in the literature. In addition, the IFT system was tested in both single-phase ‘water only’ and air-in-water transient vertical flows, for which the velocity profiles were measured with improved accuracy and temporal resolution.

One of the most significant drawbacks of modelling complex machine structures and drives using numerical models (discreet or hybrid Finite Element Analysis (FEAs)) is the difficulty of obtaining accurate modal parameters, such as stiffness and damping values of the mechanical parts as well as the accuracy of the models. Although the FEA is one of the numerical methods that are used to speed up the simulation/calculations, the dynamics of the machine tool/drives are expected to change under machining conditions, which makes numerical models inconvenient. Using Operational Modal Analysis (OMA), on-line parameters identification, can overcome the static state deviations and give more accurate results to represent the mechanical system. Thus, the project will introduce a new systematic procedure to carry out OMA on ball-screw drives. Firstly, the identification techniques are evaluated by means of simulated models and applied to identify the dynamics of the ball-screw drive using two different modelling approaches. Furthermore, this project tends to investigate the dynamics of the ball-screw driver using a novel measurement procedure to conduct OMA. The ball-screw driver of the machine is excited using Idle running to reform impulsive inertial sequences. The vibration measurements of the system were measured using a Laser Interferometer using displacement travels of the ball-screw drive. Also, the identified modal parameters of the system were compared to those captured by mounting accelerometers on the top level of the table structure using random, impulsive and cutting force excitations. The modal parameters identification was carried out by means of the improved subspace identification, which uses the auto and cross correlation of the segmented vibration signals as an input to the classical covariance subspace identification. The proposed methodology presented the ability to perform under high-sampling rates and noise suppressions. The identified results of the feed-drive system using Laser Interferometer were obtained using different feed-rates and mass weight loads to highlight the most sensitive vibration modes due to the machining process.