In recent years, Video Anomaly Detection (VAD) has shifted from conventional appearance-based modeling to semantically driven frameworks empowered by LLMs. Traditional reconstruction- and prediction-based methods, relying on motion or appearance patterns learned from normal data, often misclassify previously unseen yet semantically normal events as anomalies. To address this limitation, we propose SOR-BDNet (Semantic-Optical Representation with Boundary Detection Network), an annotation-free multimodal VAD framework that jointly leverages visual appearance and motion dynamics to generate interpretable semantic representations at the frame level. Specifically, we employ RAFT to estimate dense motion fields and concatenate the resulting flow maps with RGB images to form unified spatiotemporal inputs. These fused representations are fed into a GPT-4o-based module that generates semantic captions capturing object semantics and motion cues. Anomalies are detected by measuring semantic deviations from a memory bank constructed from normal captions. To further refine temporal boundaries, we design a boundary refinement module that integrates visual continuity constraints with contrastive feature learning based on a Swin Transformer backbone. Extensive experiments on four challenging benchmarks—UCSD-Ped2, Avenue, ShanghaiTech, and UCF-Crime—demonstrate that SOR-BDNet achieves frame-level accuracies of 97.96%, 82.86%, 87.36%, and 85.64%, respectively. These results highlight the robustness and scalability of the proposed framework, while significantly improving interpretability and generalization across diverse real-world surveillance scenarios. The source code and pretrained models are available at https://github.com/syi-coder/SOR-BDNet-Semantic-Optical-Representation-for-Boundary-Aware-Video-Anomaly-Detection-with-GPT-4o .

In recent years, the rapid advancement in video surveillance technology has significantly enhanced public safety and security. In conventional video anomaly detection approaches, there is often an exclusive focus on local information, with key temporal dynamics being overlooked. This oversight could potentially lead to a failure in recognizing dynamic anomalies, such as the sudden running of a person or the rapid movement of objects. Therefore, this study proposes a model framework structure called MTR-SCA. By utilizing widerresnet38 and Multi-Scale Temporal Relations (MTR) to capture the multi-scale temporal relationships in video time series, the framework achieves an understanding of spatial and temporal information. It introduces the Segmented Channel Attention (SCA) to enhance key information in the input feature maps and suppress less important channels for refined feature selection. We conducted experiments with the MTR-SCA network on three datasets: Avenue, ped2, and ShanghaiTech, achieving results of 97.8%, 86.8%, and 74.1% respectively.

This paper underscores the crucial role of video anonymization in smart ports, were data privacy and security hold critical importance. It introduces the Matrix Platform, specifically designed for smart environments, and highlights its strong video anonymization capabilities. The platform employs advanced Video Anonymization techniques to effectively balance the preservation of data confidentiality with the enhancement of port security. Furthermore, the paper discusses how object detection and anonymization methods are strategically employed to protect sensitive information while still allowing access to critical operational details such as cargo and vessel types. Emphasis is placed on video anonymization's pivotal role in strengthening port security by concealing high-value assets and minimizing the risk of exposing sensitive data. By integrating the Matrix Platform, ports gain the capability to proactively manage security risks, safeguard assets, and secure information. This adaptable platform can be deployed in both Edge and Cloud environments, ensuring alignment with the specific needs of smart ports, with a primary focus on data anonymization. In conclusion, as the port industry continues to evolve, this paper asserts that the adoption of video anonymization techniques is fundamental for future growth and development, providing assurance of privacy and security in this dynamic landscape.

This paper presents an approach to autonomous IoT device lifecycle management for our developed Matrix IoT platform. We discuss our approach for zero touch onboarding, IoT device failure, device end-of-life offboarding and SLAs to support device lifecycle management. We collected timings on the key stages of our proposed onboarding process. The total onboarding time takes on average 6.4 seconds to onboard a device. Therefore, when scaled to many hundreds of devices, there is a very significant time saving benefit to onboarding devices automatically, along with the benefits of reducing human error.

Linear discriminant analysis (LDA) is a powerful supervised dimensionality reduction method for analysing high-dimensional data. However, LDA cannot use locality information in data, which makes LDA degrade dramatically in performance on multimodal data. A number of LDA variants have been proposed to exploit locality information in data, including subclass-based LDAs. We discover a problem with these variants, which is that subclasses are selected on a within-class basis without considering other classes. This causes the loss of important information at class boundaries. In this paper, we present a novel variant of subclass-based LDA, Global Subclass Discriminant Analysis (GSDA). Unlike other subclass-based LDAs, GSDA selects subclasses from global clusters that may cross class boundaries, thus utilising within-class information and between-class information. More specifically, GSDA applies an effective clustering algorithm to the whole data to construct global clusters. It then utilises the local structure refining strategy on these global clusters to construct subclasses. Finally, GSDA learns a representative data subspace by maximising inter-subclass distance and minimising intra-subclass distance simultaneously. GSDA is extensively evaluated on a wide range of public datasets through comparison with the state-of-the-art LDA algorithms. Experimental results demonstrate its superiority in terms of accuracy and run times.

Abstract The work described in this paper builds upon our previous research on adoption modelling and aims to identify the best subset of features that could offer a better understanding of technology adoption. The current work is based on the analysis and fusion of two datasets that provide detailed information on background, psychosocial, and medical history of the subjects. In the process of modelling adoption, feature selection is carried out followed by empirical analysis to identify the best classification models. With a more detailed set of features including psychosocial and medical history information, the developed adoption model, using kNN algorithm, achieved a prediction accuracy of 99.41% when tested on 173 participants. The second-best algorithm built, using NN, achieved 94.08% accuracy. Both these results have improved accuracy in comparison to the best accuracy achieved (92.48%) in our previous work, based on psychosocial and self-reported health data for the same cohort. It has been found that psychosocial data is better than medical data for predicting technology adoption. However, for the best results, we should use a combination of psychosocial and medical data where it is preferable that the latter is provided from reliable medical sources, rather than self-reported.

This paper investigates architectural and capacity issues associated with analogue and digital radio over fibre fronthaul for MIMO in 5G cellular systems. The capacity of both systems is evaluated in terms of a static system deployment and also with varying traffic load. The results show that in a leased line scenario, the analogue systems offer an opportunity to reduce the cost of fibre infrastructure by more than 93% due to the more efficient use of bandwidth. but since the ARoF equipment is currently more expensive than the DRoF equivalent, there is a trade-off between CAPEX and OPEX for a given deployment.

Abstract In many real-world classification problems there exist multiple subclasses (or clusters) within a class; in other words, the underlying data distribution is within-class multimodal. One example is face recognition where a face (i.e. a class) may be presented in frontal view or side view, corresponding to different modalities. This issue has been largely ignored in the literature or at least under studied. How to address the within-class multimodality issue is still an unsolved problem. In this paper, we present an extensive study of within-class multimodality classification. This study is guided by a number of research questions, and conducted through experimentation on artificial data and real data. In addition, we establish a case for within-class multimodal classification that is characterised by the concurrent maximisation of between-class separation, between-subclass separation and within-class compactness. Extensive experimental results show that within-class multimodal classification consistently leads to significant performance gains when within-class multimodality is present in data. Furthermore, it has been found that within-class multimodal classification offers a competitive solution to face recognition under different lighting and face pose conditions. It is our opinion that the case for within-class multimodal classification is established, therefore there is a milestone to be achieved in some machine learning algorithms (e.g. Gaussian mixture model) when within-class multimodal classification, or part of it, is pursued.

We present an extension of our previous work in [1] by investigating the use of Local Septenary Patterns (LSP) for breast density classification in mammograms. The LSP operator is a variant of Local Binary Patterns (LBP) inspired by Local Ternary Patterns (LTP) and Local Quinary patterns (LQP). The main extensions in our work are i) we investigate the use of a multi-resolution technique when extracting micro texture information, ii) we investigate different neighbourhood topologies as different ways of extracting texture features, and iii) we use an additional dataset called InBreast as well as the most popular dataset in the literature, which is the Mammographic Image Analysis Society (MIAS) to further evaluate the performance of the LSP operator.

Loss function plays an important role in CNNs. However, the recent loss functions either do not apply weight and feature normalisation or do not explicitly follow the two targets of improving discriminative ability: minimising intra-class variance and maximising inter-class variance. Besides, all of them consider only the feedback information from the current mini-batch instead of the distribution information from the whole training set. In this paper, we propose a novel loss function - Global Information-based Cosine Optimal loss (Gico loss). Gico loss is applied with weight and feature normalisation, designed explicitly following the aforementioned two targets of improving discriminative ability, and is guided by the distribution information from the whole training set. Extensive experiments are conducted on multiple public datasets, which confirms the effectiveness of the proposed Gico loss and shows that we achieve state-of-the-art performance.

Softmax loss is arguably one of the most widely used loss functions in CNNs. In recent years some Softmax variants have been proposed to enhance the discriminative ability of the learned features by adding additional margin constraints, which significantly improved the state-of-the-art performance of face recognition. However, the `margin' referenced in these losses does not represent the real margin between the different classes in the training set. Furthermore, they impose a margin on all possible combinations of class pairs, which is unnecessary. In this paper we propose the Precise Adjacent Margin loss (PAM loss), which gives an accurate definition of `margin' and has precise operations appropriate for different cases. PAM loss has better geometrical interpretation than the existing margin-based losses. Extensive experiments are conducted on LFW, YTF, MegaFace and FaceScrub datasets, and results show that the proposed method has state-of-the-art performance.

Gaussian Mixture Model (GMM) is a probabilistic model for representing normally distributed subpopulations within an overall population. It is usually used for unsupervised learning to learn the subpopulations and the subpopulation assignment automatically. It is also used for supervised learning or classification to learn the boundary of subpopulations. However, the performance of GMM as a classifier is not impressive compared with other conventional classifiers such as k-nearest neighbors (KNN), support vector machine (SVM), decision tree and naive Bayes. In this paper, we attempt to address this problem. We propose a GMM classifier, SC-GMM, based on the separability criterion in order to separate the Gaussian models as much as possible. This classifier finds the optimal number of Gaussian components for each class based on the separability criterion and then determines the parameters of these Gaussian components by using the expectation maximization algorithm. Extensive experiments have been carried out on classification tasks from general data mining to face verification. Results show that SC-GMM significantly outperforms the original GMM classifier. Results also show that SC-GMM is comparable in classification accuracy to three variants of GMM classifier: Akaike Information Criterion based GMM (AIC-GMM), Bayesian Information Criterion based GMM (BIC-GMM) and variational Bayesian gaussian mixture (VBGM). However, SC-GMM is significantly more efficient than both AIC-GMM and BIC-GMM. Furthermore, compared with KNN, SVM, decision tree and naive Bayes, SC-GMM achieves competitive classification performance.

The paper presents quantitative results of a preliminary study undertaken as part of Decision Support and Information Management System for Breast Cancer (DESIREE). DESIREE is a European-funded project to improve the management of primary breast cancer through image-based, guideline-based, experience-based, and case-based information systems. In this study we explore the use of ensemble deep learning for breast mass classification in mammograms. The proposed method is based on AlexNet with some modifications in order to adapt it to our classification problem. Subsequently, model selection is performed to select the best three results based on the highest validation accuracies during the validation phase. Finally, the prediction is based on the average probability of the models. Experimental evaluation shows that accuracy from individual models ranges between 75% and 77%, but combining the best models (ensemble networks) results in over 80% classification accuracy and aura under the curve.

Computer-aided diagnosis (CAD) has great potential in providing real benefits to doctors and patients. Recent studies have, however, found lack of trust in CAD by radiologists in clinical diagnostic decision making. One of the main reasons is the lack of an appropriate confidence measure. This paper presents the first-ever study of classification confidence in the context of breast mass classification. We evaluated 11 state-of-the-art classification algorithms on breast mass image data using their confidence of classification metric, in addition to other standard evaluation metrics including accuracy and area under the curve (ROC). Experimental results show that although most classifiers produced very similar results with less than 2% difference in terms of accuracy and ROC, their performances are significantly different in terms of confidence levels. We suggest that the confidence measure should be used in conjunction with the existing performance metrics such as accuracy and ROC.

Most Information Technology (IT) services nowadays rely on one or multiple data centers. In this paper, we propose a Quality of Service (QoS) scheme for Intra/Inter data center communications systems. The proposed scheme is based on differentiation of traffic into different class categories. The differentiation is based on the source of traffic as well as the specific traffic's requirements. We conduct a comprehensive performance study for the proposed scheme. Our results demonstrate the benefits of deploying a QoS scheme in data centers especially if there is a considerable number of live virtual machine migrations among its servers. By completing such migration quickly, the Quality of Experience (QoE) will not degrade and the workload will be kept balanced among all the servers in the Data Center. Additionally, our proposed scheme considers the live virtual machine migrations among different Data Centers which is currently rapidly increasing.

Palmprint recognition has received in the last 20 years a great deal of the research community's attention. In this paper a new palmprint matching approach based on corner feature point extraction is proposed. A 72-element fixed-length descriptor is used to capture distinctive information of each feature point neighborhood and to build a measure of similarity whilst their coordinates provide a measure of proximity between the points. Matching two images takes into account both similarity and proximity measures which converts into a cost minimization problem. Our experiments carried out on a database of 250 prints from the Poly U database have yielded very good results evidenced by an EER of 0.31%.

Estimating the initial background of a scene is a key prerequisite for several applications in video analytics. In this paper, we present a simple approach that takes into account spatio-temporal motion intensities while estimating the true background. We tested the algorithm on real video sequences from the Scene Background Initialization (SBI) benchmark dataset, and the results show that the algorithm is competitive compared to the state of the art.

Breast and pectoral muscle segmentation is an essential pre-processing step for the subsequent processes in computer aided diagnosis (CAD) systems. Estimating the breast and pectoral boundaries is a difficult task especially in mammograms due to artifacts, homogeneity between the pectoral and breast regions, and low contrast along the skin-air boundary. In this paper, a breast boundary and pectoral muscle segmentation method in mammograms is proposed. For breast boundary estimation, we determine the initial breast boundary via thresholding and employ Active Contour Models without edges to search for the actual boundary. A post-processing technique is proposed to correct the overestimated boundary caused by artifacts. The pectoral muscle boundary is estimated using Canny edge detection and a pre-processing technique is proposed to remove noisy edges. Subsequently, we identify five edge features to find the edge that has the highest probability of being the initial pectoral contour and search for the actual boundary via contour growing. The segmentation results for the proposed method are compared with manual segmentations using 322, 208 and 100mammograms from the Mammographic Image Analysis Society (MIAS), INBreast and Breast Cancer Digital Repository (BCDR) databases, respectively. Experimental results show that the breast boundary and pectoral muscle estimation methods achieved dice similarity coefficients of 98.8% and 97.8% (MIAS), 98.9% and 89.6% (INBreast) and 99.2% and 91.9% (BCDR), respectively.

The emergence of social media has provided vast amounts of information that is potentially valuable for emergency management. In the EU-FP7 Project Security Systems for Language and Image Analysis (Slandail), an image analysis system has been developed to recognize the flood water images from the social media resources by incorporating with text analysis. A novel image feature descriptor has been developed to facilitate fast image processing based on incorporation of the "Squiral" (Square-Spiral) Image Processing (SIP) framework with the "Speeded-up Robust Features" (SURF). A new approach is proposed to generate an index from image recognition outcomes based on a moving window average, which presents a temporal change based on the occurrence of flooding water identified by image analysis. The evaluation for computation time and recognition were based on a batch of images obtained from the US Federal Emergency Management Agency (FEMA) media library and Facebook corpus from Germany, and the outcomes show the advantages of the proposed image features. The simulation results demonstrate the concept of the index based on a moving window average, highlighting the potential for application in emergency management.

We focus on feature extraction and selection to best represent texture and shape properties of plant diseases in an image-based leaf monitoring system implemented in a mobile-cloud environment. A number of textural and region-based features are aggregated from previous studies; also we introduce mean and peak indices of histogram-of-shape as disease property representations along with the proposed and enhanced shape features based on diseased regions. A total of 260 colour-based attributes and 163 shape attributes are searched to find the best potential features based on different aspects: probability of feature error, correlation, targeted-class relevancy and the separability quality of a feature. Experimental results show that the best selected feature set which combines colour-based and shape features yields high classification accuracy on wheat disease images captured by a smartphone camera and also provides insights into potential sets of features to be further implemented as a lightweight standalone mobile application.

In recent years, the processing of hexagonal pixel-based images has been investigated, and as a result, a number of edge detection algorithms for direct application to such image structures have been developed. We build on this paper by presenting a novel and efficient approach to the design of hexagonal image processing operators using linear basis and test functions within the finite element framework. Development of these scalable first order and Laplacian operators using this approach presents a framework both for obtaining large-scale neighborhood operators in an efficient manner and for obtaining edge maps at different scales by efficient reuse of the seven-point linear operator. We evaluate the accuracy of these proposed operators and compare the algorithmic performance using the efficient linear approach with conventional operator convolution for generating edge maps at different scale levels.

Fast image processing is a key element in achieving real-time image and video analysis. The spiral addressing scheme [10] has been an efficient tool for hexagonal image processing (HIP), whereby the image pixel indices are stored in a one-dimensional vector that enables fast processing. Unlike HIP, which requires a complex resampling scheme, we present a novel “squiral” (square spiral) image processing (SIP) framework that provides a spiral addressing scheme for direct application to standard square pixel-based images. A SIP-based non-overlapping convolution technique is developed by simulating the eye tremor phenomenon of the human visual system to accelerate computation in feature extraction. Furthermore, we deploy the proposed simulated eye tremor technique on a sequence of video frames. The preliminary results based on two action video clips demonstrate the potential of the SIP-based eye tremor model to facilitate fast video processing.

In the current work, data gleaned from an assistive technology (reminding technology), which has been evaluated with people with Dementia over a period of several years was retrospectively studied to extract the factors that contributed to successful adoption. The aim was to develop a prediction model with the capability of prospectively assessing whether the assistive technology would be suitable for persons with Dementia (and their carer), based on user characteristics, needs and perceptions. Such a prediction tool has the ability to empower a formal carer to assess, through a very limited amount of questions, whether the technology will be adopted and used.

Assistive technology has the potential to enhance the level of independence of people with dementia, thereby increasing the possibility of supporting home-based care. In general, people with dementia are reluctant to change; therefore, it is important that suitable assistive technologies are selected for them. Consequently, the development of predictive models that are able to determine a person's potential to adopt a particular technology is desirable. In this paper, a predictive adoption model for a mobile phone-based video streaming system, developed for people with dementia, is presented. Taking into consideration characteristics related to a person's ability, living arrangements, and preferences, this paper discusses the development of predictive models, which were based on a number of carefully selected data mining algorithms for classification. For each, the learning on different relevant features for technology adoption has been tested, in conjunction with handling the imbalance of available data for output classes. Given our focus on providing predictive tools that could be used and interpreted by healthcare professionals, models with ease-of-use, intuitive understanding, and clear decision making processes are preferred. Predictive models have, therefore, been evaluated on a multi-criterion basis: in terms of their prediction performance, robustness, bias with regard to two types of errors and usability. Overall, the model derived from incorporating a k-Nearest-Neighbour algorithm using seven features was found to be the optimal classifier of assistive technology adoption for people with dementia (prediction accuracy 0.84 ± 0.0242).

Mobile cloud computing aims at improving user experience through enhancing the ability of mobile applications by doing intensive tasks in the cloud. In this paper we consider an environment similar to a hybrid cloud in which the mobile device works as a private cloud. Given that the mobile phone has both limited processing resources and battery time, the proposed mobile application architecture has been designed with the capability of sending specified data/parameters to the cloud. This data is subsequently used for further processing/mining and visualization to assist in inferring further information through mapreduce. This information gives details about resource and battery consumption which will help in optimizing the relationship between the mobile device and cloud. It will also be beneficial to the optimization of the mobile application through the trends visualized in the cloud. In this paper we created an activity recognition health application as an example and helped the user about his health along with giving an insight into abnormal behavior and lifestyle trends.

This demonstration shows the integration of video analysis and search tools to facilitate the interactive retrieval of video segments depicting specific activities from surveillance footage. The implementation was developed by members of the SAVASA project for participation in the interactive surveillance event detection (SED) task of TRECVid 2012. This year, for the first time, the purpose of the interactive SED task was to evaluate systems' ability to support users in identifying video segments that depict a specific activity (event) in a large collection of surveillance video footage. Project partners worked together to analyse video and provide a query interface enabling users to search and identify matching video segments. The collaborative integration of components from multiple partners and the participation of end user partners in evaluating the system are the novel aspects of this work.

This paper presents work on integrating multiple computer vision-based approaches to surveillance video analysis to support user retrieval of video segments showing human activities. Applied computer vision using real-world surveillance video data is an extremely challenging research problem, independently of any information retrieval (IR) issues. Here we describe the issues faced in developing both generic and specific analysis tools and how they were integrated for use in the new TRECVid interactive surveillance event detection task. We present an interaction paradigm and discuss the outcomes from face-to-face end user trials and the resulting feedback on the system from both professionals, who manage surveillance video, and computer vision or machine learning experts. We propose an information retrieval approach to finding events in surveillance video rather than solely relying on traditional annotation using specifically trained classifiers.

Stroke disease incurs long periods of hospital and community care, with associated costs. Also stroke is a highly complex disease with diverse outcomes and multiple strategies and care options for therapy and care. Previously we have developed a modeling framework which classifies patients with respect to their length of stay (LOS); phase-type models can then be used to describe patient flows for each class. Also multiple outcomes, such as discharge to normal residence, nursing home, or death can be included. We here add costs to this model of the total stroke care system and determine moments for total cost of a cohort of patients and also the separate costs in different phases and stages of care. Based on stroke patients' data from the Belfast City Hospital, various scenarios are explored with a focus on comparing the costs of thrombolysis, a clot-busting therapy under different regimes.

The efficient management of networks and the provisioning of services with desired QoS guarantees is a challenge which needs to be addressed through autonomous mechanisms which are intelligent, lightweight and scalable. Recent focus on applying Machine Learning approaches to model the network and service behavioural patterns have proved to be quite effective in fulfilling the objectives of autonomous management. To this end, this paper advances on the idea of implementing a distributed management solution which harnesses the predictive capability of Bayesian Networks (BN). A multi-node distributed Call Admission Control solution (termed as BNDAC) is proposed and implemented to demonstrate the modelling and prediction power of BN. A thorough evaluation of BNDAC is presented in terms of its prediction accuracy, algorithmic complexity and decision-making speed. In an online setup, performance of BNDAC is evaluated and compared with a centralised scenario, to demonstrate its superior performance for Call Blocking Probability and QoS provisioning. Simulation results based on Opnet Modeler and Hugin Researcher show the feasibility and applicability of BNDAC solution for real-time operation and management of real world networks such as the NGN.

Learning behavioral patterns for activities of daily living in a smart home environment can be challenged by the limited number of training data that may be available. This may be due to the infrequent repetition of routine activities (e.g., once daily), the expense of using observers to label activities, and the intrusion that would be caused by the presence of observers over long time periods. It is important, therefore, to make as much use of any labeled data that are collected, however, incomplete these data may be. In this paper, we propose an algorithm for learning behavioral patterns for multi-inhabitants living in a single smart home environment, by making full use of all limited labeled activities, including incomplete data resulting from unreliable low-level sensors in this environment. Through maximum-likelihood estimation, using Expectation-Maximization, we build a model that captures both environmental uncertainties from sensor readings and user uncertainties, including variations in how individuals carry out activities. Our algorithm outperforms models that cannot handle data incompleteness, with increasing performance gains as incompleteness increases. The approach also enables the impact of particular sensors to be assessed and can thus inform sensor maintenance and deployment.

Resource allocation in the cloud is usually driven by performance predictions, such as estimates of the future incoming load to the servers or of the quality-of-service(QoS) offered by applications to end users. In this context, characterizing web workload fluctuations in an accurate way is fundamental to understand how to provision cloud resources under time-varying traffic intensities. In this paper, we investigate the Markovian Arrival Processes (MAP) and the related MAP/MAP/1 queueing model as a tool for performance prediction of servers deployed in the cloud. MAPs are a special class of Markov models used as a compact description of the time-varying characteristics of workloads. In addition, MAPs can fit heavy-tail distributions, that are common in HTTP traffic, and can be easily integrated within analytical queueing models to efficiently predict system performance without simulating. By comparison with traced riven simulation, we observe that existing techniques for MAP parameterization from HTTP log files often lead to inaccurate performance predictions. We then define a maximum likelihood method for fitting MAP parameters based on data commonly available in Apache log files, and a new technique to cope with batch arrivals, which are notoriously difficult to model accurately. Numerical experiments demonstrate the accuracy of our approach for performance prediction of web systems.

Activity recognition is an important component of patient management in smart homes where high level activities can be learned from low level sensor data. Such activity recognition utilises sensor ID, task order and time of activation to learn about patient behavior, detect anomalies and provide prompts or other interventions. In this paper we use the sensor activation times to calculate durations and then investigate several model-based clustering approaches with a view to discretising the duration data and using such data to improve activity prediction. We explore several popular approaches to characterising such duration data, namely Coxian phase type distributions and Gaussian mixture distributions. We then show how we can utilise the learned clustering components for discretisation. Finally we use simulated data, based on a real smart kitchen deployment, to compare these approaches and evaluate the discretisation results with regard to activity prediction.

With the challenge of healthcare for the increasing number of elderly people and the prevalence of chronic disease, research has been carried out on the development of assistive technologies and devices. This paper proposes a framework of context-aware physiological analysis for remote and efficient healthcare. With the relationship between the physiological function and daily activities, the online detection of abnormal situation needs to be carried out given such rich context information. Two core modules in the framework are discussed in details by proposing hierarchical online activity recognition and dynamic Cumulative Sum Control Chart (CUSUM) methods for process control. Corresponding experiments have been set up to collect both ECG data and upper-body accelerations from two healthy participants. This framework also has great potential to be used for long term health drift detection by comparison of the physiological function patterns given the activity across different periods of time.

Corner detection is used in many computer vision applications that require fast and efficient feature matching. In addition, hexagonal pixel based images have been recently investigated for image capture and processing due to their ability to represent curved structures that are common in real images better than traditional rectangular pixel based images. Therefore, we present an approach to corner detection on hexagonal images and demonstrate that accuracy is comparable to well-known existing corner detectors applied to rectangular pixel based images.

We present a biologically motivated approach to fast feature extraction on hexagonal pixel based images using the concept of eye tremor in combination with the use of the spiral architecture and convolution of non-overlapping gradient masks. We generate seven feature maps “a-trous” that can be combined into a single complete feature map, and we demonstrate that this approach is significantly faster than the use of conventional spiral convolution or the use of a neighbourhood address look-up table on hexagonal images.

The advent of IP-based Next Generation Network (NGN) and its guaranteed QoS promise has attracted significant attention from both service providers and subscribers. However, to fulfil the said promise, there is a need to provide effective Call Admission Control (CAC) based QoS provisioning solutions which are autonomous, intelligent and scalable.

Hybrid optical and wireless technology integrations have been considered as one of the most promising candidates for the next generation broadband access networks for quite some time. The integration scheme provides the bandwidth advantages of the optical networks and mobility features of the wireless networks for Subscriber Stations (SSs). It also brings economic efficiency to the network providers particularly in rural area where the existing wired telecommunication infrastructures such as Digital Subscriber Line (DSL), Cable Modem (CM), T-1/E-1 networks or fibre deployments are either costly or unreachable. For successful integration of the optical and wireless technologies there are some technical issues which need to be addressed efficiently in order to provide End-to-End (ETE) and diverse Quality of Service (QoS) for various service classes. This paper investigates the possible challenging issues for the integrated structure of the Time Division Multiplexing and Wavelength Division Multiplexing Ethernet Passive Optical Networks (TDM EPON and WDM EPON ) with the Worldwide Interoperability for Microwave Access and Wireless Fidelity (WiMAX and Wi-Fi) networks. To reduce the ETE delay and provide the QoS for diverse service classes, we have compared six existing upstream scheduling mechanisms in two levels which are distributed on Access Points (APs) from Wi-Fi domain and Base Stations (BSs) from WiMAX domain. Performance evaluations of the existing scheduling techniques for three popular service classes (Quad-play) have been studied which show the strong impact of using the efficient up-link scheduler in converged scenario. We have also proposed a dynamic scheduling algorithm for optical and wireless integration scheme, which is under the implementation and evaluation process.

Enterprise IT environments have seen a sharp growth in content use due to the popularity of on-demand data-intensive applications. In turn, the huge demand in content has spawned off major developments such as growth and distribution of computing nodes as well as the adoption of various implementation technologies. Given the complexity brought to the makeup of business computing environments in addressing the above-mentioned factors, the critical planning task of determining the appropriate infrastructure sizes for supporting firm Quality of Service (QoS) guarantees becomes a very challenging undertaking to fulfil. Benchmarking methods are widely employed in calibrating attainable performance in IT solutions, but these have the drawback of presenting output performance metrics as composite measurements that only give an end-to-end perspective. As an enhancement to benchmarking approaches, we explore the use of Performance Monitoring Counters (PMCs) in obtaining detailed operational performance of CPU and memory hardware. Performance Monitoring Counters (PMCs) are onchip registers found on most modern processor hardware. We use PMC-derived measurements to validate cache performance trends that have been derived analytically, and in the course of validations, PMC data is also used to investigate the nature and character of surges in cache miss events, which emerge as the memory load generated by runtime processes increases.

The importance of providing guaranteed Quality of Service (QoS) cannot be overemphasised, especially in the NGN environment which supports converged services on a common IP transport network. Call Admission Control (CAC) mechanisms do provide QoS to class-based services in a proactive manner. However, due to the factors of complexity, scale and dynamicity of NGN, Machine Learning techniques are favoured to analytical approaches for providing autonomous CAC. This paper is an effort to compare the performance of two such approaches - Neural Networks (NN) and Bayesian Networks (BN), to model the network behaviour and to estimate QoS metrics to be used in the CAC algorithm. It provides a way to find the optimum model training size for accurate predictions. Performance comparison is based on a wide range of experiments through a simulated network in Opnet. The outcome of this comparative study provides some interesting insights into the behaviour of NN and BN models and how they can be utilised for better CAC implementations.

A systematic design procedure is used to develop Laplacian operators that facilitate the computation of hexagonal feature map pyramids. Our focus is the development of algorithms that can operate on hexagonal images over a range of scales. We show how scalable operators can be explicitly constructed using a Gaussian neighbourhood function. We extend this approach to achieve an efficient approximation via a feature map pyramid that implicitly embodies operator scaling. In both cases we provide performance evaluation with respect to edge localisation.

We present a general approach to the computation of adaptive tri-directional operators for use on hexagonal pixel-based images, based on the spiral architecture. We show that the use of Gaussian basis functions within the finite element method provides a framework for a systematic design procedure for operators that are adaptive to spiral neighbourhoods through the use of an explicit scale parameter. We evaluate the proposed operators using simulated hexagonal images and provide comparative results with the use of traditional rectangular operators.

The Spiral Architecture has been developed as a fast way of indexing a hexagonal pixel-based image. In combination with spiral addition and spiral multiplication, methods have been developed for hexagonal image processing operations such as translation and rotation. Using the Spiral Architecture as the basis for our operator structure, we present a general approach to the computation of adaptive coarse scale Laplacian operators for use on hexagonal pixel-based images. We evaluate the proposed operators using simulated hexagonal images and demonstrate improved performance when compared with rectangular Laplacian operators such as Marr-Hildreth.

Feature extraction in image data has been investigated for many years, and more recently the problem of processing images containing irregularly distributed data has become prominent. Range data are now commonly used in the areas of image processing and computer vision. However, due to the data irregularity found in range images that occurs with a variety of image sensors, direct image processing, in particular edge detection, is a non-trivial problem. Typically, irregular range data would require to be interpolated to a regular grid prior to processing. One example of an edge detection technique than can be directly applied to range images is the scan-line approximation, but this does not employ exact data locations. Therefore, we present novel Laplacian operators that can be applied directly to irregularly distributed data, and in particular we focus on application to irregularly distributed 3D range data for the purpose of edge detection. Within the data distribution framework commonly occurring in range data acquisition devices, our results illustrate that the approach works well over a range of levels of irregularity of data distribution. The use of Laplacian operators on range data is also found to be much less susceptible to noise than the traditional use of Laplacian operators on intensity images.

Assistive technology in smart homes for elderly people with Alzheimer's disease is needed to support 'aging in place'. In this paper, we propose a probabilistic learning approach to characterise behavioural patterns for multi-inhabitants in smart homes. Decision support is then provided to monitor and assist patients to complete activities of daily living (ADL). Reasoning is based on the learned profiles and partially observed low-level sensors information. Data are stored in the proposed snow-flake schema based on homeML (an XML based schema for representation of information within smart homes). A laboratory has been developed for studying activities of 'making drinks' for multiple users. Evaluations of our learning and decision support approach are carried out on both real and simulated data. The potential of our approach to support assistive living and home-health monitoring of Alzheimer's patients is demonstrated.

Use of incomplete image data has become a prominent research issue in recent years, driven by the development of space variant image sensors. Whilst image reconstruction techniques have been developed that enable the subsequent use of standard image processing algorithms, the development of image processing algorithms that can be applied directly to incomplete image data has received less attention. The problem of interest point detection for incomplete images is addressed by presenting an algorithm that can be applied directly to incomplete image data without the requirement of image reconstruction, and the accurate performance of the algorithm is illustrated through visual results and ROC curves.

Malignant melanoma is a skin tumour typified by high mortality rates when not diagnosed and excised in its earliest stages. Preoperative diagnostic accuracy may be improved through the development of computerised systems which accurately quantify features indicative of this cancer. One such feature is boundary contour asymmetry, which is typically measured across a skin lesion's major and minor axes of symmetry. In this paper techniques for detection of skin lesion asymmetry are discussed, and the viability of integrating Fourier descriptors into a shape asymmetry quantifier is investigated. It is concluded that Fourier descriptors facilitate accurate isolation and ranking of a lesion's symmetry axes and provide an approach which could easily be integrated into new or existing diagnostic procedures.

To enable fast reliable feature matching or tracking in scenes, features need to be discrete and meaningful, and hence corner detection is often used for this purpose. However, to obtain a higher level description of an image, such as identification of objects, additional information such as edges is required, and more recently detectors have been proposed that find both edges and corners. We present a combined operator, enabling edge and corner detection to be achieved concurrently. We demonstrate that accuracy is comparable to well-known existing corner detectors and edge detectors, and, as standard post-smoothing of the corner map is not required, significantly reduced computation time can be achieved.

The use of range data has become prominent in the field of computer vision. Due to the irregular nature of range data that occurs with a number of sensors, feature extraction is a complex and challenging problem. Feature extraction techniques for range images are often based on scan line data approximations and hence do not employ exact data locations. We present a finite element based approach to the development of Laplacian operators that can be applied to both regularly or irregularly distributed range data. We demonstrate that the feature maps generated using our approach on range data are much less susceptible to noise than the traditional use of Laplacian operators on intensity images.

Range images can provide an almost 3-dimensional description of a scene. Feature driven segmentation of range images has been primarily used for 3D object recognition, and hence the accuracy of the detected features is a prominent issue. Feature extraction on range images has proven to be a more complex problem than on intensity images due to both the irregular distribution of range image data and the nature of the features that are present in range images. Approaches to range image feature extraction are often scan line based approximations that carry a significant computational overhead and hence are not appropriate for real-time processing. This paper presents a design procedure for scalable first order derivative operators that can be used directly on irregularly distributed data. Hence the method is appropriate for direct use on range image data without the requirement of image preprocessing and could form the basis of algorithms of real-time robotic applications.

The use of omnidirectional cameras has had a significant impact on the success of vision systems for video surveillance and autonomous robot navigation. Typically images obtained from such cameras are transformed to sparse panoramic images that are interpolated prior to low level image processing. We present a graph theoretic approach that enables image processing techniques, principally feature extraction, to be performed directly on sparse panoramic images, avoiding the need for image interpolation. We thus aim to reduce the computational overheads of processing images arising from omnidirectional cameras, whilst retaining accuracy sufficient for application to real-time robot vision.

Accurate segmentation of the optic nerve head (or optic disk) is very useful during the analysis and assessment of glaucoma in the eye. The Heidelberg retinal tomograph (HRT) can acquire high quality images of the optic disk and also allows three dimensional topographic measurements to be made. However, there are significant problems in optic disk segmentation due to having to deal with issues such as distractors along blood vessel edges; the extent of pallor in the optic disk or the very variable appearance of the optic nerve head itself. We propose a multi-scale region and boundary hybrid snake method to extract the optic disk. This model takes account of the vessel edge gradient direction and tries to avoid its force influence when evolving the snake. Experimental results are assessed using an overlap ratio.

The auditory brainstem response (ABR) has become a routine clinical tool for hearing and neurological assessment. In order to pick out the ABR from the background EEG activity that obscures it, stimulus-synchronized averaging of many repeated trials is necessary, typically requiring up to 2000 repetitions. This number of repetitions can be very difficult, time consuming and uncomfortable for some subjects. In this study, a method combining wavelet analysis and Bayesian networks is introduced to reduce the required number of repetitions, which could offer a great advantage in the clinical situation. 314 ABRs with 64 repetitions and 155 ABRs with 128 repetitions recorded from eight subjects are used here. A wavelet transform is applied to each of the ABRs, and the important features of the ABRs are extracted by thresholding and matching the wavelet coefficients. The significant wavelet coefficients that represent the extracted features of the ABRs are then used as the variables to build the Bayesian network for classification of the ABRs. In order to estimate the performance of this approach, stratified ten-fold cross-validation is used.

The auditory brainstem response (ABR) has become a routine clinical tool for hearing and neurological assessment. In order to pick out the ABR from the background EEG activity that obscures it, stimulus-synchronized averaging of many repeated trials is necessary and it typically requires up to 2000 repetitions. This number of repetitions can be very difficult, time consuming and uncomfortable for some subjects. In this study a method combining the wavelet analysis and the Bayesian network is introduced to reduce the required number of repetitions, which could offer a great advantage in the clinical situation. The important features of the ABR are extracted by thresholding and matching the wavelet coefficients. These extracted features are then used as the variables to build up the Bayesian network for classifying the ABR. 172 ABRs with 64 repetitions are applied in this study to learn the Bayesian network and estimate the conditional probability tables (CPTs). A further 142 ABRs with 64 repetitions are used to test the network. Moreover, this Bayesian network can also be applied to classify the ABRs with 128 repetitions.

Incomplete image data sets are of interest in many domains and arise in a variety of applications, and in particular in applications that use remote sensor array data. Although recent developments in mesh modelling of images have provided algorithms that can achieve accurate and efficient image representations without the high computational cost associated with earlier optimisation-based methods, such techniques rely on the availability of the entire image data. These content-based mesh modelling techniques aim to provide a high sample density in regions of interest, such as feature neighbourhoods or around moving objects, whilst achieving efficiency by retaining a low overall image sampling density. The sampling density is determined by a feature map, such as local image curvature or local spatial-frequency content that is obtained from the underlying complete image data. As the requirement for the availability of complete image data makes such content-based mesh modelling techniques unsuitable for application to incomplete images, where an image consists of a sparse data set, we aim to address this issue by proposing an alternative approach to mesh modelling that is based on automatically adaptive feature detection directly applicable to sparsely sampled images.

Recent developments in mesh modelling of images have provided algorithms that can achieve accurate and efficient image representations without the high computational cost associated with earlier optimisation-based methods. Hence nonuniform sampling of images combined with the use of irregular content-based meshing has provided a successful basis for recent developments in image compression techniques. The evaluation of these techniques has focussed on the accuracy and efficiency with which the mesh model can represent the image. For real-time applications, the usefulness of a mesh model may be assessed by its ability to yield compressed image representations that can be processed directly to provide output that is sufficiently accurate. Hence we present an evaluation of mesh model algorithms that is based on feature detection on the associated compressed image representations. Such an approach is built on the recent development of systematic design procedures for scalable and adaptive image processing operators that can be applied directly to non-uniformly sampled images. We demonstrate the approach using image derivative operators on compressed images.

Aggregate views are commonly used for summarizing information held in very large databases such as those encountered in data warehousing, large scale transaction management, and statistical databases. Such applications often involve distributed databases that have developed independently and therefore may exhibit incompatibility, heterogeneity, and data inconsistency. We are here concerned with the integration of aggregates that have heterogeneous classification schemes where local ontologies, in the form of such classification schemes, may be mapped onto a common ontology. In previous work, we have developed a method for the integration of such aggregates; the method previously developed is efficient, but cannot handle innate data inconsistencies that are likely to arise when a large number of databases are being integrated. In this paper, we develop an approach that can handle data inconsistencies and is thus inherently much more scalable. In our new approach, we first construct a dynamic shared ontology by analyzing the correspondence graph that relates the heterogeneous classification schemes; the aggregates are then derived by minimization of the Kullback-Leibler information divergence using the EM (Expectation-Maximization) algorithm. Thus, we may assess whether global queries on such aggregates are answerable, partially answerable, or unanswerable in advance of computing the aggregates themselves.

Non-uniformly sampled images represented on irregular content-based meshes are central to the developments in image compression techniques and in efficient motion tracking. We present a general approach to the development of systematic design procedures for scalable and adaptive low level image processing operators that can be applied to such non-uniformly sampled images. We provide algorithms that use the content-based mesh to address the usually difficult issue of local operator scale selection. The operator scale is therefore automatically matched to the local scale of the image features as embodied in the mesh. In this way we are able to apply a range of operators directly to compressed images. We demonstrate the approach with the design of image derivative operators that enable image feature detection to be implemented directly on compressed images.

The World-Wide Web provides an ever-increasing source of diverse information. We focus on query agents, in particular the matching and negotiation agents that are responsible for pre-integration where the matching agent decomposes the query into sub-queries, and then searches metadata to find datasets that match the query fragments. In the case of heterogeneous data, the matching agent utilises a negotiation agent to find datasets that match the query fragments, provides mappings from the data to the query, and constructs the appropriate (sub-)query re-writing rules. Such matching is done by generalising the data and testing if the (sub) query is matchable to the generalised (meta) data: we call this g-matchable; if it is then we can construct an operator stack to transform the data to match the (sub) query. Such an approach provides a capability of automating the process of executing queries on heterogeneous statistical databases that are distributed over the Internet. The novelty lies in the provision of automated methods for statistical aggregates, where the heterogeneity essentially resides in the classification schemes of categorical data, including both heterogeneity of nomenclature and heterogeneity of granularity. In addition, our solution permits queries to be specified in a goal-driven query-by-example format. Rather than impose an a priori global standard, the user can query through a unified interface where integration is done at run-time.

In image filtering, the 'circularity' of an operator is an important factor affecting its accuracy. For example, circular differential edge operators are effective in minimising the angular error in the estimation of image gradient direction. We present a general approach to the computation of scalable circular low-level image processing operators that is based on the finite element method. We show that the use of Gaussian basis functions within the finite element method provides a framework for a systematic and efficient design procedure for operators that are scalable to near-circular neighbourhoods through the use of an explicit scale parameter. The general design technique may be applied to a range of operators. Here we evaluate the approach for the design of an image gradient operator, and we present comparative results with other gradient approximation methods.

Information stored in a database is often subject to uncertainty and imprecision. Probability theory provides a well-known and well understood way of representing uncertainty and may thus be used to provide a mechanism for storing uncertain information in a database. We consider the problem of aggregation using an imprecise probability data model that allows us to represent imprecision by partial probabilities and uncertainty using probability distributions. Most work to date has concentrated on providing functionality for extending the relational algebra with a view to executing traditional queries on uncertain or imprecise data. However, for imprecise and uncertain data, we often require aggregation operators that provide information on patterns in the data. Thus, while traditional query processing is tuple-driven, processing of uncertain data is often attribute-driven where we use aggregation operators to discover attribute properties. The aggregation operator that we define uses the Kullback-Leibler information divergence between the aggregated probability distribution and the individual tuple values to provide a probability distribution for the domain values of an attribute or group of attributes. The provision of such aggregation operators is a central requirement in furnishing a database with the capability to perform the operations necessary for knowledge discovery in databases.

Article Discovery of multi-level rules and exceptions from a distributed database Share on Authors: Rónán Páircéir School of Information and Software Engineering, Faculty of Informatics, University of Ulster, Cromore Road, Coleraine, BT52 1SA, Northern Ireland School of Information and Software Engineering, Faculty of Informatics, University of Ulster, Cromore Road, Coleraine, BT52 1SA, Northern IrelandView Profile , Sally McClean School of Information and Software Engineering, Faculty of Informatics, University of Ulster, Cromore Road, Coleraine, BT52 1SA, Northern Ireland School of Information and Software Engineering, Faculty of Informatics, University of Ulster, Cromore Road, Coleraine, BT52 1SA, Northern IrelandView Profile , Bryan Scotney School of Information and Software Engineering, Faculty of Informatics, University of Ulster, Cromore Road, Coleraine, BT52 1SA, Northern Ireland School of Information and Software Engineering, Faculty of Informatics, University of Ulster, Cromore Road, Coleraine, BT52 1SA, Northern IrelandView Profile Authors Info & Claims KDD '00: Proceedings of the sixth ACM SIGKDD international conference on Knowledge discovery and data miningAugust 2000 Pages 523–532https://doi.org/10.1145/347090.347196Online:01 August 2000Publication History 4citation767DownloadsMetricsTotal Citations4Total Downloads767Last 12 Months1Last 6 weeks0 Get Citation AlertsNew Citation Alert added!This alert has been successfully added and will be sent to:You will be notified whenever a record that you have chosen has been cited.To manage your alert preferences, click on the button below.Manage my AlertsNew Citation Alert!Please log in to your account Save to BinderSave to BinderCreate a New BinderNameCancelCreateExport CitationPublisher SiteGet Access

It is frequently the case that data mining is carried out in an environment which contains noisy and missing data. This is particularly likely to be true when the data were originally collected for different purposes, as is commonly the case in data warehousing. In this paper we discuss the use of domain knowledge, e.g., integrity constraints or a concept hierarchy, to re-engineer the database and allocate sets to which missing or unacceptable outlying data may belong. Attribute-oriented knowledge discovery has proved to be a powerful approach for mining multi-level data in large databases. Such methods are set-oriented in that attribute values are considered to belong to subsets of the domain. These subsets may be provided directly by the database or derived from a knowledge base using inductive logic programming to re-engineer the database. In this paper we develop an algorithm which allows us to aggregate imprecise data and use it for multi-level rule induction and knowledge discovery. ©2000 John Wiley & Sons, Inc.

Distributed databases allow us to integrate data from different sources which have not previously been combined. In this article, we are concerned with the situation where the data sources are held in a distributed database. Integration of the data is then accomplished using the Dempster–Shafer representation of evidence. The weighted sum operator is developed and this operator is shown to provide an appropriate mechanism for the integration of such data. This representation is particularly suited to statistical samples which may include missing values and be held at different levels of aggregation. Missing values are incorporated into the representation to provide lower and upper probabilities for propositions of interest. The weighted sum operator facilitates combination of samples with different classification schemes. Such a capability is particularly useful for knowledge discovery when we are searching for rules within the concept hierarchy, defined in terms of probabilities or associations. By integrating information from different sources, we may thus be able to induce new rules or strengthen rules which have already been obtained. We develop a framework for describing such rules and show how we may then integrate rules at a high level without having to resort to the raw data, a useful facility for knowledge discovery where efficiency is of the essence. © 1997 John Wiley & Sons, Inc.