Egocentric Action Recognition (EAR) aims to identify fine-grained actions and interacted objects from first-person videos, forming a core task in egocentric video understanding. Despite recent progress, EAR remains challenged by limited data scale, annotation quality, and long-tailed class distributions. To address these issues, we propose REAR, a Retrieval-augmented framework for EAR that leverages external third-person (exocentric) videos as auxiliary knowledge—without requiring synchronized ego-exo pairs. REAR adopts a dual-branch architecture: one branch extracts egocentric representations, while the other retrieves semantically relevant exocentric features. These are fused via a cross-view integration module that performs staged refinement and attention-based alignment. To mitigate class imbalance, a class-adaptive selector dynamically adjusts retrieval depth based on class frequency, and independent classifiers are trained with logit-adjusted cross-entropy. Extensive experiments across three benchmarks demonstrate that REAR achieves state-of-the-art performance, with significant gains in object recognition and tail-class accuracy. The source code is publicly available at https://github.com/zou-y23/REAR.

To improve the effectiveness and efficiency of biomedical information retrieval by proposing ranking-based methods for selecting an optimal subset of retrieval systems for data fusion, we propose three ranking-based subset selection methods SFS (Sequential Forward Search), D&P (Diversity & Performance), and P&D (Performance & Diversity). These methods were applied in combination with the Reciprocal Rank Fusion technique. Experiments were conducted on four medical datasets from TREC, using between 62 and 125 candidate retrieval systems, and selecting up to 15 for fusion. The proposed subset selection methods significantly improved retrieval performance. Fusing the selected systems using RRF yielded improvements ranging from 10% to over 60% compared to the best individual retrieval system across the datasets. They also outperform the state-of-the-art technology by a large margin. In summary, our subset selection approach offers a practical and cost-efficient solution for biomedical information retrieval, achieving substantial performance gains while reducing computational overhead.

Cryptocurrency price prediction poses significant challenges due to the inherent volatility and nonlineardynamics of the market. This study introduces a hybrid stacked modeling framework that integrates machine learning (ML) and deep learning (DL) techniques, capitalizing on their complementary strengths-ML models are effective at capturing nonlinearfeature interactions in structured data, while DL architectures are adept at modeling temporal dependencies in sequential data. The proposed model leverages historical price data, technical indicators, macroeconomic variables, and sentiment metrics, with feature engineering applied to enhance predictive capability. Empirical evaluation was conducted through two experimental setups: (i) short-term, monthly segment analysis and (ii) long-term generalization via five-fold cross-validation. The hybrid model outperformed individual baseline models, achieving up to 18.3% lower RMSE and 6.7% higher directional accuracy. Additionally, it yielded superior risk-adjusted returns, with Sharpe Ratios reaching 0.094 on the Ethereum dataset. Beyond technical improvements, this research offers foresight into digital financial markets, providing a robust tool for investors, institutions, and policymakers navigating the evolving cryptocurrency landscape. The model supports more informed decision-making, enhances market oversight, and contributes to the development of adaptive regulatory frameworks for digital finance.

In egocentric videos, collecting and annotating supervised data is more complicated and time-consuming than in exocentric videos, limiting research in this area. As a remedy, Unsupervised Domain Adaptation (UDA) enhances model performance on unlabeled target domains by bridging the distribution gap between source and target domains. However, UDA for egocentric action recognition is under-explored, facing unique challenges such as simultaneous learning of verb and noun representations, focusing on human-object interactions, and managing excessive verb-noun combinations. To tackle these issues, we propose a novel Unsupervised Domain Adaptation for Egocentric Action Recognition (UDA-EAR) approach that adaptively models egocentric actions and facilitates cross-domain knowledge transfer, improving recognition performance in unlabeled target domains. Specifically, our UDA-EAR employs adaptive spatio-temporal and spatio-channel attention in a dual-branch pipeline to focus on motion intervals and interaction regions, respectively, allowing specialized learning of discriminative representations while avoiding negative combination dependencies from domain gaps. Additionally, an adversarial domain alignment mechanism aligns the data distributions between source and target domains, effectively transferring fine-grained verb-noun knowledge of egocentric videos. Extensive experiments demonstrate that our UDA-EAR outperforms state-of-the-art baselines on widely used egocentric datasets, significantly improving egocentric action recognition accuracy. Our source codes and datasets are available at https://github.com/zou-y23/UDA-EAR.

Wearable sensor-based human activity recognition (HAR) is a critical research domain in activity perception. However, achieving high efficiency and long sequence recognition remains a challenge. Despite the extensive investigation of temporal deep learning models, such as convolutional neural networks, RNNs, and transformers, their extensive parameters often pose significant computational and memory constraints, rendering them less suitable for resource-constrained mobile health applications. This study introduces HARMamba, an innovative lightweight and versatile HAR architecture that combines selective bidirectional state-space model and hardware-aware design. To optimize real-time resource consumption in practical scenarios, HARMamba employs linear recursive mechanisms and parameter discretization, allowing it to selectively focus on relevant input sequences while efficiently fusing scan and recompute operations. The model employs independent channels to process sensor data streams, dividing each channel into patches and appending classification tokens to the end of the sequence. It utilizes position embedding to represent the sequence order. The patch sequence is subsequently processed by HARMamba Block, and the classification head finally outputs the activity category. The HARMamba Block serves as the fundamental component of the HARMamba architecture, enabling the effective capture of more discriminative activity sequence features. HARMamba outperforms contemporary state-of-the-art frameworks, delivering comparable or better accuracy with significantly reducing computational and memory demands. Its effectiveness has been extensively validated on four publicly available data sets, namely, PAMAP2, WISDM, UNIMIB SHAR, and UCI. The F1 scores of HARMamba on the four data sets are 99.74%, 99.20%, 88.23%, and 97.01%, respectively.

Abstract Data fusion has demonstrated its effectiveness in enhancing information retrieval across various studies. However, advanced fusion methods typically require a dataset with extensive relevance judgments to train optimal model weights, necessitating labor-intensive and costly manual efforts. This study explores efficient methods for generating training data to facilitate affordable relevance judgments and improve fusion model quality. Experiments conducted on six datasets from TREC’s Precision Medicine and Deep Learning tracks reveal that with careful sampling design, near-optimal fusion weights can be achieved using only 5% of the documents compared to the full TREC judgments. This translates to a dataset comprising 20 queries and 500 relevance-judged documents in total. The findings highlight the potential for sophisticated fusion techniques to become more accessible to researchers and practitioners, delivering substantial performance improvements with minimal judgment effort and cost.

Learning to estimate and classify brain functional networks (BFNs) has become an increasingly important way of predicting neurological or mental disorders at their early stages. The traditional methods conduct BFN estimation and classification in two separate steps, thus preventing the interaction and joint optimization. In contrast, Transformer provides a natural architecture to learn BFNs with downstream tasks in an end-to-end manner. Despite their great potential, Transformer-based methods involve a large number of parameters that need to be learnt from Big Data and often lead to poor model interpretability. Considering the challenge in acquiring data and the high demand for model interpretability in medical scenarios, in this paper, we propose a minimalist Transformer architecture, referred to as Miniformer, by simplifying the projection matrices in the self-attention module into a single diagonal matrix, which greatly reduces the number of parameters, alleviates the risk of overfitting, and improves the interpretability. Additionally, the clear physical meaning of parameters in Miniformer makes the integration of domain knowledge or prior easier and more natural. Therefore, we further develop two variants of Miniformer by incorporating sparsity for removing potentially noisy time points from fMRI signals, and smoothness for capturing the temporal correlations in fMRI signals, respectively. To evaluate the effectiveness of the proposed methods, we perform brain disease diagnosis experiments on three public datasets. The results show that Miniformer and its variants tend to achieve higher classification performance than comparison methods with good interpretability.

Smart environment is an efficient and cost-effective way to afford intelligent supports for the elderly people. Human activity recognition is a crucial aspect of the research field of smart environments, and it has attracted widespread attention lately. The goal of this study is to develop an effective sensor-based human activity recognition model based on the belief-rule-based system (BRBS), which is one of representative rule-based expert systems. Specially, a new belief rule base (BRB) modeling approach is proposed by taking into account the self- organizing rule generation method and the multi-temporal rule representation scheme, in order to address the problem of combination explosion that existed in the traditional BRB modelling procedure and the time correlation found in continuous sensor data in chronological order. The new BRB modeling approach is so called self-organizing and multi-temporal BRB (SOMT-BRB) modeling procedure. A case study is further deducted to validate the effectiveness of the SOMT-BRB modeling procedure. By comparing with some conventional BRBSs and classical activity recognition models, the results show a significant improvement of the BRBS in terms of the number of belief rules, modelling efficiency, and activity recognition accuracy.

Abstract Mobile notifications have become ubiquitous in modern life, yet excessive volumes contribute to interruption overload. This paper investigates intelligent notification management leveraging user context. A three-stage methodology employed a focus group, survey, and in-the-wild data collection app. The focus group $$(n=12)$$ ( n = 12 ) provided preliminary insights into notification perceptions during boredom which informed survey design. The survey $$(n=106)$$ ( n = 106 ) probed usage habits across times, days, and app categories. The SeektheNotification app gathered real-world notification data from 20 Android users over 3 months.Analysis revealed social and personal apps dominate notification volumes (91% combined). Shorter response delays occurred on weekends and after 12 pm, suggesting heightened user receptivity during boredom. Random Forest classification achieved 88% accuracy, outperforming 13 other algorithms, underscoring machine learning’s potential for context-aware notification systems.Our exploratory findings indicate notifications could be optimized by considering situational factors like boredom. Further research should expand context beyond boredom and employ advanced deep learning techniques. This preliminary study demonstrates the promise of leveraging user psychology and machine intelligence to develop smarter interruption management systems to combat notification overload.

Currently, Human Activity Recognition (HAR) applications need a large volume of data to be able to generalize to new users and environments. However, the availability of labeled data is usually limited and the process of recording new data is costly and time-consuming. Synthetically increasing datasets using Generative Adversarial Networks (GANs) has been proposed, outperforming cropping, time-warping, and jittering techniques on raw signals. Incorporating GAN-generated synthetic data into datasets has been demonstrated to improve the accuracy of trained models. Regardless, currently, there is no optimal GAN architecture to generate accelerometry signals, neither a proper evaluation methodology to assess signal quality or accuracy using synthetic data. This work is the first to propose conditional Wasserstein Generative Adversarial Networks (cWGANs) to generate synthetic HAR accelerometry signals. Furthermore, we calculate quality metrics from the literature and study the impact of synthetic data on a large HAR dataset involving 395 users. Results show that i) cWGAN outperforms original Conditional Generative Adversarial Networks (cGANs), being 1D convolutional layers appropriate for generating accelerometry signals, ii) the performance improvement incorporating synthetic data is more significant as the dataset size is smaller, and iii) the quantity of synthetic data required is inversely proportional to the quantity of real data.

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.

Traditional clustering algorithms are widely used for building bag-of-words (BOW) models to aggregate spatio-temporal feature points extracted from a video for human activity recognition problems. Their performances are restricted by the computational complexity which limits the number of feature points being used. In contrast, deep clustering yields good clustering performance without the limit of the number of feature points. Therefore, this work proposes a dual stacked autoencoders features embedded clustering (DSAFEC) and a BOW construction method based on the DSAFEC (B-DSAFEC) to reduce the computational complexity and to remove the selection restriction. The DSAFEC first transforms feature points extracted from a video to a learned feature space and then probabilities of cluster assignment of feature points are predicted to build BOWs for human activity recognition. A soft clustering is used by assigning each feature point to multiple clusters yielding the largest probabilities instead of only one in hard clustering. Experimental results on three benchmark human activity datasets show that the B-DSAFEC yields better performance compared to five reference methods which are developed based on either traditional clustering methods or deep clustering methods.

With the development of the cloud-based Internet of Things (IoT), people and things can request services, access data, or control actuators located thousands of miles away. The entity authentication of the remotely accessed devices is an essential part of the security systems. In this vein, physical unclonable functions (PUFs) are a hot research topic, especially for generating random, stable, and tamper-resistant fingerprints. This article proposes a lightweight, robust static random access memory (SRAM)-PUF-based entity authentication scheme to guarantee that the accessed end devices are trustable. The proposed scheme uses challenge-response pairs (CRPs) represented by reordered memory addresses as challenges and the corresponding SRAM cells' startup values as responses. The experimental results show that our scheme can efficiently authenticate resources-constrained IoT devices with a low computation overhead and small memory capacity. Furthermore, we analyze the SRAM-PUF by testing the PUF output under different environmental conditions, including temperature and magnetic field, in addition to exploring the effect of writing different values to the SRAM cells on the stability of their startup values.

The Synthetic Minority Oversampling Technique (SMOTE) is effective to handle imbalance classification problems. However, the random candidate selection of SMOTE may lead to severe overlap between classes and introduce new noise factors. Many variants of SMOTE have been proposed to relieve these problems by generating new examples in safe regions. Most of these methods generate new examples with existing minority examples without considering the negative impact that class imbalance have brought on these examples. In this paper, we handle the imbalance classification using Bayes' decision rule and propose a novel oversampling method, the Minority Oversampling using Sensitivity (MOSS). Candidates for new example generations are selected considering their sensitivity with respect to class imbalance. New examples are then generated by interpolating the candidate and one of its adjacent examples. Experiments on 30 datasets confirm the superiority of the MOSS against one baseline method and seven oversampling methods.

The Internet of things is a rapidly expanding paradigm which is fundamentally altering the way in which we interact with technology. A range of new services are enabled by this technological revolution, one of which is the task of activity recognition from performing classification on sensed information, which is an area of research related to assisted living. The ability to reliably sense the environment is a crucial aspect of this application area, which has, to date, been prone to error and poor data quality. Therefore, it is essential that we are able to identify potentially anomalous data before it can have a serious effect in the application domain and evoke dangerous consequences. This work presents a novel Markov-based technique for detecting anomalous sensor events in constrained Internet of Things environments. Results from the experiment found a recall score of 97.9% and an F-measure of 92.0%, which represents a promising step forward in this research direction.

Abstract The use of Convolutional Neural Networks (CNNs) as a feature learning method for Human Activity Recognition (HAR) is becoming more and more common. Unlike conventional machine learning methods, which require domain-specific expertise, CNNs can extract features automatically. On the other hand, CNNs require a training phase, making them prone to the cold-start problem. In this work, a case study is presented where the use of a pre-trained CNN feature extractor is evaluated under realistic conditions. The case study consists of two main steps: (1) different topologies and parameters are assessed to identify the best candidate models for HAR, thus obtaining a pre-trained CNN model. The pre-trained model (2) is then employed as feature extractor evaluating its use with a large scale real-world dataset. Two CNN applications were considered: Inertial Measurement Unit (IMU) and audio based HAR. For the IMU data, balanced accuracy was 91.98% on the UCI-HAR dataset, and 67.51% on the real-world Extrasensory dataset. For the audio data, the balanced accuracy was 92.30% on the DCASE 2017 dataset, and 35.24% on the Extrasensory dataset.

Abstract The Internet of Things (IoT) is rapidly changing the way in which we engage with technology on a daily basis. The IoT paradigm enables low-resource devices to intercommunicate in a fully flexible and pervasive manner, and the data from these devices is used for decision-making in critical applications such as; traffic infrastructure, health-care and home security, to name but a few. Due to the scarce resources available in these IoT devices, being able to quantify the reliability of them is a critical function. This report presents a detailed evolution of the area of reliability measurement, followed by an in-depth review of the state-of-the-art for quantification of reliability in the IoT, revealing the many challenges associated with this task. From this in-depth review, a set of key research directions for IoT reliability is determined. Despite the critical nature of the research area, at this current moment, this study is the first detailed review available in the area of assessing IoT reliability.

The articles in this special section provide a snapshot of the latest research advancements in all aspects of connected health and informatic systems where artificial intelligence has been evident, including sensing, transfer, storage and analytics of biomedical data. These articles capture the end-to-end view of solutions that use automated informatics to address single ormultiple scenarios of health engineering such as primary care, preventive care, predictive technologies, hospitalization, home care, and occupational health.

Abstract Current approaches to patient education provide generic standardized materials to all patients regardless of their demographics such as age and cognitive abilities. Thus, the effectiveness of this approach may suffer from a patient's motivation to fully engage with the material. To alleviate these concerns, this study proposes a personalized approach to patient education that is tailored to the individual characteristics and health objectives of the patient. Personalized features will enhance the comprehensibility and usability of the process of medical education. Taking this into consideration, this paper introduces a conceptual architecture to create a web‐based personalized patient education experience. A key component of this architecture comprises ontological models of the patient themselves, their medical conditions, physical activities and their educational attainments. Furthermore, rule‐based reasoning is also proposed to achieve this personalization. A use case scenario is provided to highlight the effectiveness of personalized education provision.

The global population is aging; projections show that by 2050, more than 20% of the population will be aged over 64. This will lead to an increase in aging related illness, a decrease in informal support, and ultimately issues with providing care for these individuals. Assistive smart homes provide a promising solution to some of these issues. Nevertheless, they currently have issues hindering their adoption. To help address some of these issues, this study introduces a novel approach to implementing assistive smart homes. The devised approach is based upon an intention recognition mechanism incorporated into an intelligent agent architecture. This approach is detailed and evaluated. Evaluation was performed across three scenarios. Scenario 1 involved a web interface, focusing on testing the intention recognition mechanism. Scenarios 2 and 3 involved retrofitting a home with sensors and providing assistance with activities over a period of 3 months. The average accuracy for these three scenarios was 100%, 64.4%, and 83.3%, respectively. Future will extend and further evaluate this approach by implementing advanced sensor-filtering rules and evaluating more complex activities.

Situation awareness may be inferred from user context such as body posture transition and location data. Smartphones and smart homes incorporate sensors that can record this information without significant inconvenience to the user. Algorithms were developed to classify activity postures to infer current situations; and to measure user's physical location, in order to provide context that assists such interpretation. Location was detected using a subarea-mapping algorithm; activity classification was performed using a hierarchical algorithm with backward reasoning; and falls were detected using fused multiple contexts (current posture, posture transition, location, and heart rate) based on two models: “certain fall” and “possible fall.” The approaches were evaluated on nine volunteers using a smartphone, which provided accelerometer and orientation data, and a radio frequency identification network deployed at an indoor environment. Experimental results illustrated falls detection sensitivity of 94.7% and specificity of 85.7%. By providing appropriate context the robustness of situation recognition algorithms can be enhanced.

The accurate detection of changes has the potential to form a fundamental component of systems which autonomously solicit user interaction based on transitions within an input stream, for example, electrocardiogram data or accelerometry obtained from a mobile device. This solicited interaction may be utilized for diverse scenarios such as responding to changes in a patient's vital signs within a medical domain or requesting user activity labels for generating real-world labelled datasets. Within this paper, we extend our previous work on the Multivariate Online Change detection Algorithm subsequently exploring the utility of incorporating the Benjamini Hochberg method of correcting for multiple comparisons. Furthermore, we evaluate our approach against similarly light-weight Multivariate Exponentially Weighted Moving Average and Cumulative Sum based techniques. Results are presented based on manually labelled change points in accelerometry data captured using 10 participants. Each participant performed nine distinct activities for a total period of 35 minutes. The results subsequently demonstrate the practical potential of our approach from both accuracy and computational perspectives.

Activity Monitoring is a key feature of health and well-being assessment that has received increased consideration from the research community over the last few decades. Body worn sensors and smart devices are widely used in Activity Monitoring in order to capture and classify large amounts of data over short periods of time, in a relatively un-obtrusive manner. Change point detection is a technique at the core of the data processing of the sensory data recorded used to identify the transition from one underlying time series generation model to another. The sudden change in mean, variance or both may represent change point in time series data. Accurate and automatic change point detection in data is not only used to identify events (transition from one activity to another), however, can also be used for labelling activities to generate real world annotated datasets. This paper proposes a genetic algorithm (GA) that identifies the optimal set of parameters for a Multivariate Exponentially Weighted Moving Average (MEWMA) approach to change point detection. The proposed technique optimizes different parameters of the MEWMA in an effort to find the maximum F-measure, which subsequently identifies the exact location of the change point from an existing activity to a new one. Results have been evaluated based on real and synthetic datasets collected from accelerometer data during a set of 8 different activities for two users with a high degree of accuracy form 99.4% to 99.8% and F-measure to 66.7%.

The development, testing and evaluation of novel approaches to Intelligent Environment data processing require access to datasets which are of high quality, validated and annotated. Access to such datasets is limited due to issues including cost, flexibility, practicality, and a lack of a globally standardized data format. These limitations are detrimental to the progress of research. This paper provides an overview of the Open Data Initiative and the use of simulation software (IE Sim) to provide a platform for the objective assessment and comparison of activity recognition solutions. To demonstrate the approach, a dataset was generated and distributed to 3 international research organizations. Results from this study demonstrate that the approach is capable of providing a platform for benchmarking and comparison of novel approaches.

A smart environment is a physical space where smart devices/technologies are applied to continuously sense the occupant's daily living activities or health condition. Recent years have witnessed the application of smart environments to support independent living for elderly people or for people with chronic conditions. Nevertheless, most of the projects have been exclusively designed to support instances of single occupancy. In an effort to move beyond the scenario of a single occupant within a smart environment, this paper investigates the feasibility of using thermal sensors to detect the presence of multi-occupants and to track and visualise their motions. Results indicate that the use of thermals sensor can detect multi-occupancy, including moving subjects in addition to static subjects. The system developed demonstrated the ability to track up to 5 occupants, although mistrackings were found to occur when the individuals were apart after they stayed closely together.

In today's world of internet connected devices and smart phones, it has become effortless to create and consume vast amounts of information. This is particularly the case with photographs, with vast amounts being created and shared online every day. Never-the-less, it still remains a challenge to discover the "right" information for the appropriate purpose. This paper describes and discusses the testing and evaluation of the MediaPlace system with the well-known dataset YFCC-100M, which contains 48 million geospatial geotagged photographs, from Flickr produced by Yahoo. MediaPlace is a system which we have developed to automatically enrich geotagged photographs with semantic geospatial information derived from several online geospatial datasets.

In spite of the importance of activity recognition (AR) for intelligent human-computer interaction in emerging smart space applications, state-of-the-art AR technology is not ready or adequate for real-world deployments due to its insufficient accuracy. The accuracy limitation is directly attributed to uncertainties stemming from multiple sources in the AR system. Hence, one of the major goals of AR research is to improve system accuracy by minimizing or managing the uncertainties encountered throughout the AR process. As we cannot manage uncertainties well without measuring them, we must first quantify their impact. Nevertheless, such a quantification process is very challenging given that uncertainties come from diverse and heterogeneous sources. In this article, we propose an approach, which can account for multiple uncertainty sources and assess their impact on AR systems. We introduce several metrics to quantify the various uncertainties and their impact. We then conduct a quantitative impact analysis of uncertainties utilizing data collected from actual smart spaces that we have instrumented. The analysis is intended to serve as groundwork for developing “diagnostic” accuracy measures of AR systems capable of pinpointing the sources of accuracy loss. This is to be contrasted with the currently used accuracy measures.

A dramatic shift in population ageing is taking place, which is subsequently leading to a rise in the prevalence of chronic conditions. COPD is one such condition that places a heavy burned on national Health Services. A healthier lifestyle can greatly reduce the speed of conditional decline in those with COPD and therefore provide a better quality of life. This paper investigates how technology can be utilized for lifestyle support through self-management providing participants with the ability to monitor information on their lifestyle, behavior and physiological parameters. This information is analyzed and presented to the participant providing a comprehensive picture of their health and wellbeing. With this information at hand, informed lifestyle decisions and behavioral changes can be made and lead to a better quality of life.

In this paper we present a design, development and evaluation of a tool to facilitate the reminiscence of life-logged data intended for persons with dementia. Using off-the-shelf technologies such as a smartphone it is possible to effectively record and log a person's daily activates such as places visited and persons interacted with. We developed a reminiscence tool that visualizes life-logging data. The system was evaluated with six healthy participants aged between 24-46 years of age. They recorded approximately 2.5 hours of life-logging data that was later loaded into the developed tool. During the subsequent evaluation, the participant's gaze was tracked using eye-tracking hardware. Results show that the tool was easy to use and navigate through, facilitating the objective identification of events that may be useful for reminiscence.

Self-management is an approach to healthcare which aims to empower individuals to manage their own health conditions. This is of particular importance given the shifting demographics, increased prevalence of chronic conditions and financial austerity facing many countries. In this paper we present our current work on the development of a flexible, generic self-management platform which can be readily extended for specific chronic conditions. A Unified Modelling Language (UML) system class diagram is presented with the class design based upon functional requirements derived from literature and engagement with key stakeholders. We subsequently extend the UML class diagram to demonstrate how the design may be adapted for specific conditions.

In this paper we outline the design, development and evaluation of a smartphone based life-logging and social interaction reminder system intended for use by persons with dementia. By using a smartphone, the wearer's daily activities can be recorded in picture format, along with meta data providing activity levels and location data. In addition to this data, social interactions can also be logged and subsequently identified, using Quick Response (QR) codes. The intervention was evaluated on six healthy participants aged between 24 - 46 years of age who wore the system for 2.5 hours. The qualitative feedback received was that the technology was easy to use and was responsive and accurate at identifying, recording and displaying social interaction data.

Electrocardiograms can be used for diagnosing various cardiac conditions. They are traditionally printed as a hard copy on thermal graph paper, which a clinician can then use to support the overall diagnosis. Nevertheless, as technology evolves aspects of healthcare are embracing these technological advancements. One such technology that is being accepted is the Smartphone. Lightweight, portable and with impressive processing speed, smartphones are becoming a familiar sight within healthcare, and used not only for the purposes of making phone calls. There are many healthcare applications available which are already proving popular amongst clinicians and patients, such as WebMD, Trusted Health and Wellness Information as well as First Aid - Emergency Handbook. Additionally, with a range of five to twelve mega pixel camera as standard, images can be taken and details can be presented in ways previously not thought possible. This paper reports upon the motivations of using a smartphone within healthcare and describes the findings of having fifteen electrocardiograms diagnoses using an Apple iPhone by a clinician.

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.

Activity models play a critical role for activity recognition and assistance in ambient assisted living. Existing approaches to activity modeling suffer from a number of problems, e.g., cold-start, model reusability, and incompleteness. In an effort to address these problems, we introduce an ontology-based hybrid approach to activity modeling that combines domain knowledge based model specification and data-driven model learning. Central to the approach is an iterative process that begins with “seed” activity models created by ontological engineering. The “seed” models are deployed, and subsequently evolved through incremental activity discovery and model update. While our previous work has detailed ontological activity modeling and activity recognition, this paper focuses on the systematic hybrid approach and associated methods and inference rules for learning new activities and user activity profiles. The approach has been implemented in a feature-rich assistive living system. Analysis of the experiments conducted has been undertaken in an effort to test and evaluate the activity learning algorithms and associated mechanisms.

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).

One of the most popular techniques of generating classifier ensembles is known as stacking which is based on a meta-learning approach. In this paper, we introduce an alternative method to stacking which is based on cluster analysis. Similar to stacking, instances from a validation set are initially classified by all base classifiers. The output of each classifier is subsequently considered as a new attribute of the instance. Following this, a validation set is divided into clusters according to the new attributes and a small subset of the original attributes of the instances. For each cluster, we find its centroid and calculate its class label. The collection of centroids is considered as a meta-classifier. Experimental results show that the new method outperformed all benchmark methods, namely Majority Voting, Stacking J48, Stacking LR, AdaBoost J48, and Random Forest, in 12 out of 22 data sets. The proposed method has two advantageous properties: it is very robust to relatively small training sets and it can be applied in semi-supervised learning problems. We provide a theoretical investigation regarding the proposed method. This demonstrates that for the method to be successful, the base classifiers applied in the ensemble should have greater than 50% accuracy levels.

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 paper introduces an ontology-based integrated framework for activity modeling, activity recognition and activity model evolution. Central to the framework is ontological activity modeling and semantic-based activity recognition, which is supported by an iterative process that incrementally improves the completeness and accuracy of activity models. In addition, the paper presents a service-oriented architecture for the realization of the proposed framework which can provide activity context-aware services in a scalable distributed manner. The paper further describes and discusses the implementation and testing experience of the framework and services in the context of smart home based assistive living.

Improvements and portability of technologies and smart devices has enabled rapid growth in the amount of user generated media such as photographs and videos. Whilst various media generation and management systems exist it still remains a challenge to discover the "right" metadata information for the right purpose. This paper describes an approach to extract relevant geospatial information through reverse geocoding in addition to cross-referencing several public geospatial data sources. The extracted geospatial information can be used to enable enrichment of media with rich semantic geo-metadata and therefore enable improved organisation and searching of the media. Central to the system is the cross-referencing and data fusion of several public geospatial datasets to determine the most relevant Points of Interest and extract their relevant features. These relevant Points of Interest and features are subsequently used to annotate media with human readable information, leading to enriched media repositories. The system has been implemented as a client/server architecture, with a web interface for the client front end and Java for the backend server side processing. Details of the implementation are discussed. Testing in a scenario has been undertaken and a discussion of the testing technique and results is presented. The initial results show the system to be effective at fusing several public geospatial datasets and extracting relevant geo-metadata.

Automated monitoring and the recognition of activities of daily living (ADLs) is a key challenge in ambient-assisted living (AAL) for the assistance of the elderly. Within this context, a formal approach may provide a means to fill the gap between the low-level observations acquired by sensing devices and the high-level concepts that are required for the recognition of human activities. We describe a system named ARA (Automated Recognizer of ADLs) that exploits propositional temporal logic and model checking to support automated real-time recognition of ADLs within a smart environment. The logic is shown to be expressive enough for the specification of realistic patterns of ADLs in terms of basic actions detected by a sensorized environment. The online model checking engine is shown to be capable of processing a stream of detected actions in real time. The effectiveness and viability of the approach are evaluated within the context of a smart kitchen, where different types of ADLs are repeatedly performed.

To provide robust healthcare services and personalized recommendations details relating to a patient's daily life activities, profile information, and personal experience is of vital importance. This paper focuses on improvement in general health status of elderly patients through the use of an innovative service which align dietary intake with activity information. Personalized healthcare services based on the patient's activities of daily living and their shared experience, are provided as outputs. A knowledge driven approach has been used where all the daily life activities, social interactions, and profile information are modeled in an ontology. The semantic context is exploited that enables fine-grained situation analysis for recommendation of personalized services and decision support. Preliminary experimental results for the dynamic nature of the systems and its corresponding personalized recommendations have been found to be encouraging.

The combination of multiple classifiers, commonly referred to as an ensemble, has previously demonstrated the ability to improve overall classification accuracy in many application domains. Some ensemble techniques, however, cannot easily improve the performance of stable classification methods. One such example of a stable classification method is the k Nearest Neighbor (kNN) Classifier. In this paper we propose an alternative to the kNN ensemble method through the use of a clustering technique applied for the purpose of selecting the neighborhood of a new instance. In addition, a novel combination function based on exponential support (ExSupp) has been introduced. The proposed approach exhibited improved classification results in 16 out 20 data sets which were considered in comparison with a single kNN and a kNN ensemble based approach. Besides higher classification accuracy the proposed method exhibited higher levels of efficiency in terms of classification time.

Characterization of the kinetic and conformational properties of channel proteins is a crucial element in the integrative study of congenital cardiac diseases. The proteins of the ion channels of cardiomyocytes represent an important family of biological components determining the physiology of the heart. Some computational studies aiming to understand the mechanisms of the ion channels of cardiomyocytes have concentrated on Markovian stochastic approaches. Mathematically, these approaches employ Chapman-Kolmogorov equations coupled with partial differential equations. As the scale and complexity of such subcellular and cellular models increases, the balance between efficiency and accuracy of algorithms becomes critical. We have developed a novel two-stage splitting algorithm to address efficiency and accuracy issues arising in such modeling and simulation scenarios. Numerical experiments were performed based on the incorporation of our newly developed conformational kinetic model for the rapid delayed rectifier potassium channel into the dynamic models of human ventricular myocytes. Our results show that the new algorithm significantly outperforms commonly adopted adaptive Runge-Kutta methods. Furthermore, our parallel simulations with coupled algorithms for multicellular cardiac tissue demonstrate a high linearity in the speedup of large-scale cardiac simulations.

Current clinical methods for the assessment of Parkinson's disease suffer from inconvenience, infrequency and subjectivity. WiiPD is an approach for the objective home based assessment of Parkinson's disease which utilizes the intuitive and sensor rich Nintendo Wii Remote. Combined with an electronic patient diary, a suite of mini-games, a metric analyzer, and a visualization engine, we propose that this system can complement existing clinical practice by providing objective metrics gathered frequently over extended periods of time. In this paper we detail the approach and introduce a series of metrics deemed capable of quantifying the severity of tremor and bradykinesia in those with Parkinson's disease. The system has been tested on a 71 year old participant with Parkinson's disease over a period of 15 days, a 72 year old control user without Parkinson's disease, and a group of 8 young adults. Results indicate a clear correlation between patient self rating scores of tremor severity and metric values obtained, in addition to clear differences in metrics obtained from each user group. These results suggest that this approach is capable of indicating the presence and severity of the motor symptoms of Parkinson's disease that affect arm motor control.

This paper introduces a knowledge-driven approach to real-time, continuous activity recognition based on multisensor data streams in smart homes. The approach goes beyond the traditional data-centric methods for activity recognition in three ways. First, it makes extensive use of domain knowledge in the life cycle of activity recognition. Second, it uses ontologies for explicit context and activity modeling and representation. Third and finally, it exploits semantic reasoning and classification for activity inferencing, thus enabling both coarse-grained and fine-grained activity recognition. In this paper, we analyze the characteristics of smart homes and Activities of Daily Living (ADL) upon which we built both context and ADL ontologies. We present a generic system architecture for the proposed knowledge-driven approach and describe the underlying ontology-based recognition process. Special emphasis is placed on semantic subsumption reasoning algorithms for activity recognition. The proposed approach has been implemented in a function-rich software system, which was deployed in a smart home research laboratory. We evaluated the proposed approach and the developed system through extensive experiments involving a number of various ADL use scenarios. An average activity recognition rate of 94.44 percent was achieved and the average recognition runtime per recognition operation was measured as 2.5 seconds.

Research on sensor-based activity recognition has, recently, made significant progress and is attracting growing attention in a number of disciplines and application domains. However, there is a lack of high-level overview on this topic that can inform related communities of the research state of the art. In this paper, we present a comprehensive survey to examine the development and current status of various aspects of sensor-based activity recognition. We first discuss the general rationale and distinctions of vision-based and sensor-based activity recognition. Then, we review the major approaches and methods associated with sensor-based activity monitoring, modeling, and recognition from which strengths and weaknesses of those approaches are highlighted. We make a primary distinction in this paper between data-driven and knowledge-driven approaches, and use this distinction to structure our survey. We also discuss some promising directions for future research.

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.

Pervasive Healthcare Systems (PHS) constitute a research field that examines a wide range of technologies for the development of healthcare applications. Due to the data-centric nature of such applications, a number of challenges exist, notably combining the data generated from heterogeneous Distributed Sensor/Actuator Networks (DSANs) and maintaining application software as underlying technologies evolve. Coupling the technological features of DSANs with healthcare application software results in every sensor upgrade requiring a corresponding software upgrade, leading to constrained interoperability and degraded efficiency for the PHS. Consequently, a requirement exists to decouple technological features of DSANs operating within pervasive healthcare spaces from the corresponding healthcare applications. In this paper, the Sen-sorMed software library is introduced. SensorMed aims to address the issues related to the provision of homogeneous access to heterogeneous DSANs in an open and interoperable manner. Details of the architecture, highlighting interactions that promote the interoperability and portability of SensorMed, are presented.

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.

In this work we adopt a previously introduced meta-learning classification method for semi-supervised learning problems. In our previous work we illustrated that the method is successful when applied in a supervised classification problem. In our current work the results demonstrate that following refinements made to the method it can be successfully applied to semi-supervised classification cases.

This paper introduces a new weight factor method for lattice-based evidential fusion for the purposes of activity recognition within smart environments. In calculating the weight factor between the lattice layers, the uncertainty information derived from sensors along with the sensor context has been taken into consideration. According to the experimental results, the proposed weight factor method has the ability to effectively incorporate the uncertainty into the inference process, and subsequently infer complex activities such as a preparing lunch activity with an accuracy of 65.20%.

As a clinically complex neurodegenerative disease, Parkinson's disease (PD) requires regular assessment and close monitoring. In our current study, we have developed a home-based tool designed to monitor and assess peripheral motor symptoms. An evaluation of the tool was carried out over a period of ten weeks on ten people with idiopathic PD. Participants were asked to use the tool twice daily over four days, once when their medication was working at its best ("on" state) and once when it had worn off ("off" state). Results showed the ability of the data collected to distinguish the "on" and "off" state and also demonstrated statistically significant differences in timed assessments. It is anticipated that this tool could be used in the home environment as an early alert to a change in clinical condition or to monitor the effects of changes in prescribed medications used to manage PD.

Chronic pain is a common long-term condition that affects a person's physical and emotional functioning. Currently, the integrated biopsychosocial approach is the mainstay treatment for people with chronic pain. Self-reporting (the use of questionnaires) is one of the most common methods to evaluate treatment outcome. The questionnaires can consist of more than 300 questions, which is tedious for people to complete at home. This paper presents a machine learning approach to analyze self-reporting data collected from the integrated biopsychosocial treatment, in order to identify an optimal set of features for supporting self-management. In addition, a classification model is proposed to differentiate the treatment stages. Four different feature selection methods were applied to rank the questions. In addition, four supervised learning classifiers were used to investigate the relationships between the numbers of questions and classification performance. There were no significant differences between the feature ranking methods for each classifier in overall classification accuracy or AUC ( p > 0.05); however, there were significant differences between the classifiers for each ranking method ( p < 0.001). The results showed the multilayer perceptron classifier had the best classification performance on an optimized subset of questions, which consisted of ten questions. Its overall classification accuracy and AUC were 100% and 1, respectively.

This paper explores a sensor fusion method applied within smart homes used for the purposes of monitoring human activities in addition to managing uncertainty in sensor-based readings. A three-layer lattice structure has been proposed, which can be used to combine the mass functions derived from sensors along with sensor context. The proposed model can be used to infer activities. Following evaluation of the proposed methodology it has been demonstrated that the Dempster-Shafer theory of evidence can incorporate the uncertainty derived from the sensor errors and the sensor context and subsequently infer the activity using the proposed lattice structure. The results from this study show that this method can detect a toileting activity within a smart home environment with an accuracy of 88.2%.

Intelligent Environments on varying scales and for different purposes are slowly becoming a reality. In the near future, global smart world infrastructures will become a commodity that will support various activities of daily life at different degrees of realism. Such infrastructures have the potential to offer dedicated, context- and situation-aware information and services by simultaneously providing the next-generation of data collection, execution and service provisioning layers. One key aspect of this vision is the correct monitoring and understanding of how people interact with their environment; how they can actually benefit from the added intelligence; and finally how future services can be improved or better personalized to enhance human environment interaction as a whole. This level of intelligence is of particular relevance in the health and social care domain where person-centric services can be deployed to assist or even enable a person in performing activities of daily living. This paper discusses the concept of embedded self-aware profiles for smart devices that can be used to gain a deeper contextual understanding of their use and also discusses the emergence of a general model of Ambient Intelligence that is based on the collective existence and behavior of such smart devices. Although generic in principle, the proposed concepts have been exemplified by a distinct use case, namely a smart kettle.

This paper explores an improvement to activity recognition within a Smart Home environment using the Dempster-Shafer theory of evidence. This approach has the ability to be used to monitor human activities in addition to managing uncertainty in sensor based readings. A three layer lattice structure has been proposed, which can be used to combine the mass functions derived from sensors along with sensor context and subsequently can be used to infer activities. From the total 209 recorded activities throughout a two week period, 85 toileting activities were considered. The results from this work demonstrated that this method was capable of detecting 75 of the toileting activities correctly within a Smart Home environment equating to a classification accuracy of 88.2%.

The number of intelligent environment implementations such as smart homes is set to increase dramatically within the next 40 years. This is predicted using forecasts of demographic data which indicates an expansion of the aged population. It has also been predicted that governments will struggle to meet the demand for resources such as sensor technology due to costs. Optimisation of limited resources involves physically positioning devices to maximise pertinent data gathering potential. Currently the most utilised methodology of distributing limited spatial detection sensors such as pressure mats within smart homes is via ad-hoc deployments performed by a human being. In this study idiosyncratic inhabitant spatial-frequency data was processed using a Pure Random Search (PRS) algorithm to uncover probabilistic future regions of interest, alluding to optimal sensor distributions under resource constraint. With PRS a null hypothesis was stated: `using lower iteration stopping criteria produce less optimal sensor distributions than when using higher iteration stopping criteria'. A student t-test between 1000 and 5000 iterations was statistically significant at 5% (p = 0.016852) whereby the null hypothesis was rejected. Similar results were obtained between other iteration criteria. These data demonstrate that the iteration stopping criterion is not as critical as sensor size or number of sensors; and that comparable results could be obtained when lower stopping parameters are specified when using PRS.

This paper presents details of a convenient and unobtrusive system for monitoring daily activities. A smart phone equipped with an embedded 3D-accelerometer was worn on the belt for the purposes of data recording. Once collected the data was processed to identify 6 activities offline (walking, posture transition, gentle motion, standing, sitting and lying). The processing technique adopted a novel hierarchical classification. In the first instance, rule-based reasoning is used to discriminate between motion and motionless activities. Following this the classification process utilizes two multiclass SVM (support vector machines) classifiers to classify the motion and motionless activities, respectively. The classifiers were trained on data from one subject and tested on 10 subjects. The experiments demonstrate that the hierarchical method can reduce misclassification between motion and motionless activities. The average accuracy was improved compared with using a single classifier by using this classification method (82.8% vs. 63.8%), and is important for providing appropriate feedback in free living applications.

Smart spaces not only are surrounded by networked computers, mobile devices, and ubiquitous sensors, but also encapsulate a large amount of social and communication information. Humans, the center ...

Smart spaces are not only surrounded by networked computers, mobile devices, and ubiquitous sensors, they encapsulate a large amount of social and communication information. Humans, the center of p...

There is currently much interest in exploring new ways to optimize ECG acquisition. In the current study, we have investigated optimal configurations of ECG leads with respect to: 1) best signal magnitude (maximal signal variance) and 2) best reconstruction of the total body surface potential distribution and the 12-lead ECG. Principal component analysis was applied to a set of 117-lead body surface potential maps (BSPMs) recorded from 559 subjects. Three bipolar leads, referred to as "eigenleads," were identified from the extrema on the resulting eigenvectors. Recording sites for the three leads were largely located in the precordial region. The magnitude of the signals recorded from the eigenleads was calculated on a set of 185 unseen subjects. The accuracy of the eigenleads in the reconstruction of BSPMs and the 12-lead ECG was also assessed for each subject. These results were compared to existing limited lead systems. It was found that, when compared to conventional leads, eigenleads could be used to increase signal strength (rms voltage) by 27.9%, 39.0%, and 20.3% for P-waves, QRS segments, and STT segments, respectively. Although the eigenleads were not able to reconstruct total body surface information as well as the 12-lead ECG (24.4 mu V versus 20.2 mu V), the eigenleads did perform comparably with other limited lead systems in the estimation of the 12-lead ECG. In particular, the eigenleads performed well in the reconstruction of precordial leads in comparison to the EASI lead system and a limited lead system made up of a subset of precordial leads. The proposed leads are a suitable alternative limited leads system, and can be used to improve SNR. More work is needed to test the practicality of such leads.

New medical devices and services enabling citizenpsilas health care anywhere and anytime are expected in the near future. However, to become a reality these devices must be supported by personalized services which satisfy user needs. In this paper we propose a general approach to manage the complexity of ambiguously related information for providing enhanced, user-specific services in the self-care domain. The global architecture is driven by a compositional model between a domain-specific and a context-awareness model, which aggregates the citizenpsilas and the devices profiles, the citizenpsilas healthcare characteristics and available signal processing methods. The final objective is to support automatic composition of services helping any citizen to select an optimal and personalized sensor system and to improve decision-making.

Smart Homes (SH) have emerged as a realistically viable solution capable of providing technology-driven assistive living for the elderly and disabled. Nevertheless, it still remains a challenge to provide situation-aware assistance for those in need in their Activity of Daily Living (ADL). This paper introduces a systematic approach to providing situation-aware ADL assistances in a smart home environment. The approach makes use of semantic technologies for sensor data modeling, fusion and management, thus creating machine understandable and processable situational data. It exploits intelligent agents for interpreting and reasoning semantic situational (meta)data to enhance situation-aware decision support. We analyze the nature and issues of SH-based healthcare for cognitively deficient inhabitants. We discuss the ways in which semantic technologies enhance situation comprehension. We describe a cognitive agent for realizing high-level cognitive capabilities such as prediction and explanation. We outline the implementation of a prototype assistive system and illustrate the proposed approach through simulated and real-time ADL assistance scenarios in the context of situation aware assistive living.

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.

Body surface potential maps were investigated to identify a set of optimal recording sites required to discriminate between several diseases. Specifically, recordings captured from subjects exhibiting myocardial infarction or left ventricular hypertrophy, as well as a control group consisting of healthy subjects, were investigated. Owing to the fact that multi-class problems are inherently difficult to solve we divided the problem into several two-class scenarios. Six data sets were generated from the available 744 records, each viewing the available data differently, to form several two-class problems. A data-driven selection algorithm was applied to each of the generated data sets to produce six classification models, each utilizing as features those recording sites offering most to the discrimination task being investigated. Subsequently, a framework was introduced to facilitate the combination of outputs from each classifier. Essentially, the framework used the outputs from half of the classification models to determine which of the remaining models would be employed to form a final decision. A benchmark, in the form of a multi-group classifier, was introduced to evaluate the perceived benefits of the proposed approach. An improvement of approximately 10% upon the benchmark was observed resulting in an overall accuracy of 79.19%.

This special section editorial defines home automation and investigates the various approaches to home automation which can be used to facilitate independent living. The paper concludes by introducing the three technical papers that are part of the special section.

Advances in wearable health systems, from a smart textile, signal processing, and wireless communications perspective, have resulted in the recent deployment of such systems in real clinical and healthcare settings. Nevertheless, the problem of identifying the most appropriate sites from which biological parameters can be recorded still remains unsolved. This paper aims to asses the effects of various practical constraints that may be encountered when choosing electrocardiographic recording sites for wearable health systems falling within the category of smart shirts for cardiac monitoring and analysis. We apply a lead selection algorithm to a set of 192 lead body surface potential maps (BSPM) and simulate a number of practical constraints by only allowing selection of recording sites from specific regions available in the 192 lead array. Of the various scenarios that were investigated, we achieved the best results when the selection process to identify the recording sites was constrained to an area around the precordial region. The top ten recording sites chosen in this region exhibited an rms voltage error of 25.8 mu V when they were used to estimate total ECG information. The poorest performing scenario was that which constrained the selection to two vertical strips on the posterior surface. The top ten recording sites chosen in this scenario exhibited an rms voltage error of 41.1 muV. In general, it was observed that out of all the scenarios investigated, those which constrained available regions to the posterior and lateral surfaces performed less favorably than those where electrodes could also be chosen on the anterior surface. The overall results from our approach have validated the proposed algorithm and its ability to select optimal recording sites taking into consideration the practical constraints that may exist with smart shirts.

This article presents the conceptual communication model of dynamic groups, that dynamically utilizes three traditional communication metaphors through the use of context-based information. Dynamic groups makes creation, management and usage of groups easy. It enables social network structures to be maintained in both virtual and face-to-face settings as well as in the combination thereof. This article defines the dynamic management of advanced contact lists which can include presence and status information, a/synchronous multimedia communication tools, and methods for structuring social networks. It also contains an initial evaluation and a proposed architecture for technical realisation.

In this article, we investigate four variations (D-HSM, D-HSW, D-HSE, and D-HSEW) of a novel indexing technique called D-HS designed for use in case-based reasoning (CBR) systems. All D-HS modifications are based on a matrix of cases indexed by their discretized attribute values. The main differences between them are in their attribute discretization stratagem and similarity determination metric. D-HSM uses a fixed number of intervals and simple intersection as a similarity metric; D-HSW uses the same discretization approach and a weighted intersection; D-HSE uses information gain to define the intervals and simple intersection as similarity metric; D-HSEW is a combination of D-HSE and D-HSW. Benefits of using D-HS include ease of case and similarity knowledge maintenance, simplicity, accuracy, and speed in comparison to conventional approaches widely used in CBR. We present results from the analysis of 20 case bases for classification problems and 15 case bases for regression problems. We demonstrate the improvements in accuracy and/or efficiency of each D-HS modification in comparison to traditional k-NN, R-tree, C4,5, and M5 techniques and show it to be a very attractive approach for indexing case bases. We also illuminate potential areas for further improvement of the D-HS approach. © 2007 Wiley Periodicals, Inc. Int J Int Syst 22: 353–383, 2007.

Autonomic computing provides mechanisms for the self-management of computing systems. This paper proposes context prediction as an autonomic mechanism to improve the usability of ubiquitous computing environments; in our case, for behavioral prediction in an environment that supports independent living for ageing people. A purpose of this paper is to stimulate debate on how best to improve ubiquitous computing environments for those who inhabit them; in particular for those ageing people who wish to continue to live independently. We propose a layered and extensible context architecture that provides self-managed and self-configuration capabilities. In particular, in addition to the context provider and context service layers, we propose context prediction and context fusion layers with cross-layer context-quality capabilities. We envisage a distributed peer-to-peer architecture with some form of semantic overlay network facilitating autonomic communications within and beyond the ubiquitous home environment.

Body surface potential mapping (BSPM) is a technique employing multiple electrodes to capture, via noninvasive means, an indication of the heart's condition. An inherent problem with this technique is the resulting high-dimensional recordings and the subsequent problems for diagnostic classifiers. A data set, recorded from a 192-lead BSPM system, containing 74 records is investigated. QRS isointegral maps, offering a summary of the information obtained during ventricular depolarization, were derived from 30 old inferior myocardial infarction and 44 normal recordings. Principal component analysis was applied to reduce the dimensionality of the recordings and a linear classifier was employed for classification. This perceptron-based classifier has been adapted so that the final weight and bias values are estimated prior to the learning process. This estimation process, referred to as the linear hyperplane approach (LHA), derives the estimated weights from a bisector hyperplane, placed orthogonal to the means of two class distributions in an n-dimensional Euclidean space. Estimating weights encourages a network to exhibit better generalization ability. Utilizing a number of different principal components as input features, the LHA achieved an average sensitivity and specificity of 79.58% and 76.45%, respectively, across all experiments. The average accuracy of 76.73% achieved with this approach was significantly better than the other benchmark classifiers evaluated against it.

In this paper, we present a modified perceptron that has been optimised to enhance its generalization capabilities through pre-initialisation of the synaptic weights. The rationale for the research is presented highlighting the obvious benefits of such an approach. A description of results obtained from an experiment seeking cardiac classification is presented. The dataset used contained 74 patient records; 30 inferior myocardial infarction and 44 normal. Patient records where acquired using 192-lead body surface potential maps (BSPM). QRS isointegral maps where derived from the 192 lead maps before applying principal component analysis to reduce the dimensionality of the dataset. Only the first 2 principal components were used to create a 2 dimensional space, which can be both easily visualized and modeled by a single perceptron. In addition to the optimised approach proposed, several other classification methods were evaluated on the dataset to generate benchmarks from which comparisons were made. These included linear, probabilistic, non-linear and neural network methods. The optimised approach resulted in sensitivity and specificity figures of 73.33% and 70.45% respectively and provided an overall accuracy which was 2.5% higher than that of the next best classifier.

We investigate an algorithmic extension to the technique of stacked regression that prunes the size of a homogeneous ensemble set based on a consideration of the accuracy and diversity of the set members. We show that the pruned ensemble set is as accurate on average over the data-sets tested as the nonpruned version, which provides benefits in terms of its application efficiency and reduced complexity of the ensemble.