Data augmentation for Human Activity Recognition with Generative Adversarial Networks.

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.

Evaluating the effectiveness of mindfulness alone compared to exercise and mindfulness on fatigue in women with gynaecology cancer (GEMS): Protocol for a randomised feasibility trial.

BACKGROUND: In 2020 Globocan reported nearly 1.4 million new cases of gynaecology cancer worldwide. Cancer related fatigue has been identified as a symptom that can be present for gynaecology cancer patients many years after treatment. The current evidence around the management of this symptom suggests that exercise has the most positive outcome. However, some ambiguity remains around the evidence and whether it can address all areas of fatigue effectively. More recently, other interventions such as mindfulness have begun to show a favourable response to the management of symptoms for cancer patients. To date there has been little research that explores the feasibility of using both these interventions together in a gynaecology cancer population. This study aims to explore the feasibility of delivering an intervention that involves mindfulness and mindfulness and exercise and will explore the effect of this on fatigue, sleep, mood and quality of life. METHODS/DESIGN: This randomised control trial will assess the interventions outcomes using a pre and post design and will also include a qualitative process evaluation. Participants will be randomised into one of 2 groups. One group will undertake mindfulness only and the other group will complete exercise and mindfulness. Both groups will use a mobile application to complete these interventions over 8 weeks. The mobile app will be tailored to reflect the group the participants have drawn during randomisation. Self-reported questionnaire data will be assessed at baseline prior to commencing intervention and at post intervention. Feasibility will be assessed through recruitment, adherence, retention and attrition. Acceptability and participant perspective of participation (process evaluation), will be explored using focus groups. DISCUSSION: This trial will hope to evidence and demonstrate that combination of two interventions such as mindfulness and exercise will further improve outcomes of fatigue and wellbeing in gynaecology cancer. The results of this study will be used to assess (i) the feasibility to deliver this type of intervention to this population of cancer patients using a digital platform; (ii) assist this group of women diagnosed with cancer to manage fatigue and other symptoms of sleep, mood and impact their quality of life. TRIAL REGISTRATION: NCT05561413.

Predicting Activity Duration in Smart Sensing Environments Using Synthetic Data and Partial Least Squares Regression: The Case of Dementia Patients.

The accurate recognition of activities is fundamental for following up on the health progress of people with dementia (PwD), thereby supporting subsequent diagnosis and treatments. When monitoring the activities of daily living (ADLs), it is feasible to detect behaviour patterns, parse out the disease evolution, and consequently provide effective and timely assistance. However, this task is affected by uncertainties derived from the differences in smart home configurations and the way in which each person undertakes the ADLs. One adjacent pathway is to train a supervised classification algorithm using large-sized datasets; nonetheless, obtaining real-world data is costly and characterized by a challenging recruiting research process. The resulting activity data is then small and may not capture each person's intrinsic properties. Simulation approaches have risen as an alternative efficient choice, but synthetic data can be significantly dissimilar compared to real data. Hence, this paper proposes the application of Partial Least Squares Regression (PLSR) to approximate the real activity duration of various ADLs based on synthetic observations. First, the real activity duration of each ADL is initially contrasted with the one derived from an intelligent environment simulator. Following this, different PLSR models were evaluated for estimating real activity duration based on synthetic variables. A case study including eight ADLs was considered to validate the proposed approach. The results revealed that simulated and real observations are significantly different in some ADLs (p-value < 0.05), nevertheless synthetic variables can be further modified to predict the real activity duration with high accuracy (R2(pred)>90%).

Modelling mobile-based technology adoption among people with dementia.

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.

A Systematic Multidisciplinary Process for User Engagement and Sensor Evaluation: Development of a Digital Toolkit for Assessment of Movement in Children With Cerebral Palsy.

Objectives: To describe and critique a systematic multidisciplinary approach to user engagement, and selection and evaluation of sensor technologies for development of a sensor-based Digital Toolkit for assessment of movement in children with cerebral palsy (CP). Methods: A sequential process was employed comprising three steps: Step 1: define user requirements, by identifying domains of interest; Step 2: map domains of interest to potential sensor technologies; and Step 3: evaluate and select appropriate sensors to be incorporated into the Digital Toolkit. The process employed a combination of principles from frameworks based in either healthcare or technology design. Results: A broad range of domains were ranked as important by clinicians, patients and families, and industry users. These directly informed the device selection and evaluation process that resulted in three sensor-based technologies being agreed for inclusion in the Digital Toolkit, for use in a future research study. Conclusion: This report demonstrates a systematic approach to user engagement and device selection and evaluation during the development of a sensor-based solution to a healthcare problem. It also provides a narrative on the benefits of employing a multidisciplinary approach throughout the process. This work uses previous frameworks for evaluating sensor technologies and expands on the methods used for user engagement.

Design and Implementation of a Smart Home in a Box to Monitor the Wellbeing of Residents With Dementia in Care Homes.

There is a global challenge related to the increasing number of People with Dementia (PwD) and the diminishing capacity of governments, health systems, and caregivers to provide the best care for them. Cost-effective technology solutions that enable and ensure a good quality of life for PwD via monitoring and interventions have been investigated comprehensively in the literature. The objective of this study was to investigate the challenges with the design and deployment of a Smart Home In a Box (SHIB) approach to monitoring PwD wellbeing within a care home. This could then support future SHIB implementations to have an adequate and prompt deployment allowing research to focus on the data collection and analysis aspects. An important consideration was that most care homes do not have the appropriate infrastructure for installing and using ambient sensors. The SHIB was evaluated via installation in the rooms of PwD with varying degrees of dementia at Kirk House Care Home in Belfast. Sensors from the SHIB were installed to test their capabilities for detecting Activities of Daily Living (ADLs). The sensors used were: (i) thermal sensors, (ii) contact sensors, (iii) Passive Infrared (PIR) sensors, and (iv) audio level sensors. Data from the sensors were collected, stored, and handled using a 'SensorCentral' data platform. The results of this study highlight challenges and opportunities that should be considered when designing and implementing a SHIB approach in a dementia care home. Lessons learned from this investigation are presented in addition to recommendations that could support monitoring the wellbeing of PwD. The main findings of this study are: (i) most care home buildings were not originally designed to appropriately install ambient sensors, and (ii) installation of SHIB sensors should be adapted depending on the specific case of the care home where they will be installed. It was acknowledged that in addition to care homes, the homes of PwD were also not designed for an appropriate integration with ambient sensors. This study provided the community with useful lessons, that will continue to be applied to improve future implementations of the SHIB approach.

Leveraging Wearable Sensors for Human Daily Activity Recognition with Stacked Denoising Autoencoders.

Activity recognition has received considerable attention in many research fields, such as industrial and healthcare fields. However, many researches about activity recognition have focused on static activities and dynamic activities in current literature, while, the transitional activities, such as stand-to-sit and sit-to-stand, are more difficult to recognize than both of them. Consider that it may be important in real applications. Thus, a novel framework is proposed in this paper to recognize static activities, dynamic activities, and transitional activities by utilizing stacked denoising autoencoders (SDAE), which is able to extract features automatically as a deep learning model rather than utilize manual features extracted by conventional machine learning methods. Moreover, the resampling technique (random oversampling) is used to improve problem of unbalanced samples due to relatively short duration characteristic of transitional activity. The experiment protocol is designed to collect twelve daily activities (three types) by using wearable sensors from 10 adults in smart lab of Ulster University, the experiment results show the significant performance on transitional activity recognition and achieve the overall accuracy of 94.88% on three types of activities. The results obtained by comparing with other methods and performances on other three public datasets verify the feasibility and priority of our framework. This paper also explores the effect of multiple sensors (accelerometer and gyroscope) to determine the optimal combination for activity recognition.

SemImput: Bridging Semantic Imputation with Deep Learning for Complex Human Activity Recognition.

The recognition of activities of daily living (ADL) in smart environments is a well-known and an important research area, which presents the real-time state of humans in pervasive computing. The process of recognizing human activities generally involves deploying a set of obtrusive and unobtrusive sensors, pre-processing the raw data, and building classification models using machine learning (ML) algorithms. Integrating data from multiple sensors is a challenging task due to dynamic nature of data sources. This is further complicated due to semantic and syntactic differences in these data sources. These differences become even more complex if the data generated is imperfect, which ultimately has a direct impact on its usefulness in yielding an accurate classifier. In this study, we propose a semantic imputation framework to improve the quality of sensor data using ontology-based semantic similarity learning. This is achieved by identifying semantic correlations among sensor events through SPARQL queries, and by performing a time-series longitudinal imputation. Furthermore, we applied deep learning (DL) based artificial neural network (ANN) on public datasets to demonstrate the applicability and validity of the proposed approach. The results showed a higher accuracy with semantically imputed datasets using ANN. We also presented a detailed comparative analysis, comparing the results with the state-of-the-art from the literature. We found that our semantic imputed datasets improved the classification accuracy with 95.78% as a higher one thus proving the effectiveness and robustness of learned models.

Evaluating the Impact of a Two-Stage Multivariate Data Cleansing Approach to Improve to the Performance of Machine Learning Classifiers: A Case Study in Human Activity Recognition.

Human activity recognition (HAR) is a popular field of study. The outcomes of the projects in this area have the potential to impact on the quality of life of people with conditions such as dementia. HAR is focused primarily on applying machine learning classifiers on data from low level sensors such as accelerometers. The performance of these classifiers can be improved through an adequate training process. In order to improve the training process, multivariate outlier detection was used in order to improve the quality of data in the training set and, subsequently, performance of the classifier. The impact of the technique was evaluated with KNN and random forest (RF) classifiers. In the case of KNN, the performance of the classifier was improved from 55.9% to 63.59%.

Semi-Automated Data Labeling for Activity Recognition in Pervasive Healthcare.

Activity recognition, a key component in pervasive healthcare monitoring, relies on classification algorithms that require labeled data of individuals performing the activity of interest to train accurate models. Labeling data can be performed in a lab setting where an individual enacts the activity under controlled conditions. The ubiquity of mobile and wearable sensors allows the collection of large datasets from individuals performing activities in naturalistic conditions. Gathering accurate data labels for activity recognition is typically an expensive and time-consuming process. In this paper we present two novel approaches for semi-automated online data labeling performed by the individual executing the activity of interest. The approaches have been designed to address two of the limitations of self-annotation: (i) The burden on the user performing and annotating the activity, and (ii) the lack of accuracy due to the user labeling the data minutes or hours after the completion of an activity. The first approach is based on the recognition of subtle finger gestures performed in response to a data-labeling query. The second approach focuses on labeling activities that have an auditory manifestation and uses a classifier to have an initial estimation of the activity, and a conversational agent to ask the participant for clarification or for additional data. Both approaches are described, evaluated in controlled experiments to assess their feasibility and their advantages and limitations are discussed. Results show that while both studies have limitations, they achieve 80% to 90% precision.

The Value of Including People with Dementia in the Co-Design of Personalized eHealth Technologies.

BACKGROUND: In this article, we discuss the benefits and implications of the shift from a user-centered to a co-creation approach in the processes of designing and developing eHealth and mHealth solutions for people with dementia. To this end, we illustrate the case study of a participatory design experience, implemented at the REMIND EU Project, Connected Health Summer School, which took place in June 2018 at Artimino (Italy). OBJECTIVES: The initiative was intended to reach two objectives: (1) help researchers specializing in a variety of fields (engineering, computing, psychology, nursing, and dementia care) develop a deeper understanding of how individuals living with dementia expect to be supported and/or enabled by eHealth and mHealth technologies and (2) prevent the tendency to focus on the impairments that characterize dementia at the expense of seeing the individual living with this condition as a whole person, striving to maintain a life that is as fulfilling as possible. METHOD: The Connected Health Summer School is an annual multidisciplinary training program, organized in collaboration with the REMIND EU Project, designed for early-stage researchers interested in the development of new eHealth and mHealth services and apps. For the 2018 program edition, REMIND end user partner Novilunio invited two members of the Irish Dementia Working Group to deliver keynote lectures, and engage in participatory workshops to facilitate the creation of digital technology applications based on their specific real-life needs, values, and expectations. Their involvement as participants and experts was aimed to give a clear message to early-stage researchers: a true personalized approach to eHealth and mHealth solutions can only emerge from a highly reflective and immersive appreciation of people's subjective accounts of their lived experience. RESULTS/CONCLUSIONS: The Connected Health Summer School early-stage researchers developed 6 app mock-ups based on their discussions and co-creation activities with the two experts with dementia. The reflections on this experience highlight a number of important issues that demand consideration when undertaking eHealth and mHealth research, co-design, and development with and for people with dementia. The evolution in design research from a user-centered approach to co-designing should pave the way to the development of technologies that neither disempower nor reinforce stigma, but instead provide a reliable support to living a life as active and meaningful as possible after a diagnosis of dementia. To this end, the motto of the peak global organization of people with dementia, Dementia Alliance International, says it all: "See the person and not the dementia."

A Thermal Imaging Solution for Early Detection of Pre-ulcerative Diabetic Hotspots.

Foot ulcers are a common complication of diabetes and are the leading cause of amputation amongst those with diabetes. Research has shown that, an increase of two degrees Celsius in the skin temperature on the plantar surface of the foot can be an early indication of injury or inflammation. Early detection and treatment of a hotspot region may reduce the risk of an ulcer developing. This paper presents a thermography-based approach for detecting temperature hotspots on the foot. The system comprises a bespoke application and a thermal camera attachment which captures RGB images and a temperature matrix. Web-based services process the captured data and detect whether any regions of higher temperature are present on the foot, in comparison to the other foot. The accuracy of this system has been verified through a pilot study. Hotspots were simulated on the feet of 10 healthy participants. The results indicated that hotspots were correctly detected for 60% of the participants. We discuss some reasons why the results were inaccurate for the remaining four participants. Furthermore, we also suggest some potential enhancements to the system with the aim of increasing the precision of the results.

Automatic Annotation for Human Activity Recognition in Free Living Using a Smartphone.

Data annotation is a time-consuming process posing major limitations to the development of Human Activity Recognition (HAR) systems. The availability of a large amount of labeled data is required for supervised Machine Learning (ML) approaches, especially in the case of online and personalized approaches requiring user specific datasets to be labeled. The availability of such datasets has the potential to help address common problems of smartphone-based HAR, such as inter-person variability. In this work, we present (i) an automatic labeling method facilitating the collection of labeled datasets in free-living conditions using the smartphone, and (ii) we investigate the robustness of common supervised classification approaches under instances of noisy data. We evaluated the results with a dataset consisting of 38 days of manually labeled data collected in free living. The comparison between the manually and the automatically labeled ground truth demonstrated that it was possible to obtain labels automatically with an 80⁻85% average precision rate. Results obtained also show how a supervised approach trained using automatically generated labels achieved an 84% f-score (using Neural Networks and Random Forests); however, results also demonstrated how the presence of label noise could lower the f-score up to 64⁻74% depending on the classification approach (Nearest Centroid and Multi-Class Support Vector Machine).

Ethical adoption: A new imperative in the development of technology for dementia.

INTRODUCTION: Technology interventions are showing promise to assist persons with dementia and their carers. However, low adoption rates for these technologies and ethical considerations have impeded the realization of their full potential. METHODS: Building on recent evidence and an iterative framework development process, we propose the concept of "ethical adoption": the deep integration of ethical principles into the design, development, deployment, and usage of technology. RESULTS: Ethical adoption is founded on five pillars, supported by empirical evidence: (1) inclusive participatory design; (2) emotional alignment; (3) adoption modelling; (4) ethical standards assessment; and (5) education and training. To close the gap between adoption research, ethics and practice, we propose a set of 18 practical recommendations based on these ethical adoption pillars. DISCUSSION: Through the implementation of these recommendations, researchers and technology developers alike will benefit from evidence-informed guidance to ensure their solution is adopted in a way that maximizes the benefits to people with dementia and their carers while minimizing possible harm.

Modelling assistive technology adoption for people with dementia.

PURPOSE: Assistive technologies have been identified as a potential solution for the provision of elderly care. Such technologies have in general the capacity to enhance the quality of life and increase the level of independence among their users. Nevertheless, the acceptance of these technologies is crucial to their success. Generally speaking, the elderly are not well-disposed to technologies and have limited experience; these factors contribute towards limiting the widespread acceptance of technology. It is therefore important to evaluate the potential success of technologies prior to their deployment. MATERIALS AND METHODS: The research described in this paper builds upon our previous work on modelling adoption of assistive technology, in the form of cognitive prosthetics such as reminder apps and aims at identifying a refined sub-set of features which offer improved accuracy in predicting technology adoption. Consequently, in this paper, an adoption model is built using a set of features extracted from a user's background to minimise the likelihood of non-adoption. The work is based on analysis of data from the Cache County Study on Memory and Aging (CCSMA) with 31 features covering a range of age, gender, education and details of health condition. In the process of modelling adoption, feature selection and feature reduction is carried out followed by identifying the best classification models. FINDINGS: With the reduced set of labelled features the technology adoption model built achieved an average prediction accuracy of 92.48% when tested on 173 participants. CONCLUSIONS: We conclude that modelling user adoption from a range of parameters such as physical, environmental and social perspectives is beneficial in recommending a technology to a particular user based on their profile.

The Empowering Role of Mobile Apps in Behavior Change Interventions: The Gray Matters Randomized Controlled Trial.

BACKGROUND: Health education and behavior change programs targeting specific risk factors have demonstrated their effectiveness in reducing the development of future diseases. Alzheimer disease (AD) shares many of the same risk factors, most of which can be addressed via behavior change. It is therefore theorized that a behavior change intervention targeting these risk factors would likely result in favorable rates of AD prevention. OBJECTIVE: The objective of this study was to reduce the future risk of developing AD, while in the short term promoting vascular health, through behavior change. METHODS: The study was an interventional randomized controlled trial consisting of subjects who were randomly assigned into either treatment (n=102) or control group (n=42). Outcome measures included various blood-based biomarkers, anthropometric measures, and behaviors related to AD risk. The treatment group was provided with a bespoke "Gray Matters" mobile phone app designed to encourage and facilitate behavior change. The app presented evidence-based educational material relating to AD risk and prevention strategies, facilitated self-reporting of behaviors across 6 behavioral domains, and presented feedback on the user's performance, calculated from reported behaviors against recommended guidelines. RESULTS: This paper explores the rationale for a mobile phone-led intervention and details the app's effect on behavior change and subsequent clinical outcomes. Via the app, the average participant submitted 7.3 (SD 3.2) behavioral logs/day (n=122,719). Analysis of these logs against primary outcome measures revealed that participants who improved their high-density lipoprotein cholesterol levels during the study duration answered a statistically significant higher number of questions per day (mean 8.30, SD 2.29) than those with no improvement (mean 6.52, SD 3.612), t97.74=-3.051, P=.003. Participants who decreased their body mass index (BMI) performed significantly better in attaining their recommended daily goals (mean 56.21 SD 30.4%) than those who increased their BMI (mean 40.12 SD 29.1%), t80 = -2.449, P=.017. In total, 69.2% (n=18) of those who achieved a mean performance percentage of 60% or higher, across all domains, reduced their BMI during the study, whereas 60.7% (n=34) who did not, increased their BMI. One-way analysis of variance of systolic blood pressure category changes showed a significant correlation between reported efforts to reduce stress and category change as a whole, P=.035. An exit survey highlighted that respondents (n=83) reported that the app motivated them to perform physical activity (85.4%) and make healthier food choices (87.5%). CONCLUSIONS: In this study, the ubiquitous nature of the mobile phone excelled as a delivery platform for the intervention, enabling the dissemination of educational intervention material while simultaneously monitoring and encouraging positive behavior change, resulting in desirable clinical effects. Sustained effort to maintain the achieved behaviors is expected to mitigate future AD risk. TRIAL REGISTRATION: ClinicalTrails.gov NCT02290912; https://clinicaltrials.gov/ct2/show/NCT02290912 (Archived by WebCite at http://www.webcitation.org/6ictUEwnm).

Dynamic detection of window starting positions and its implementation within an activity recognition framework.

Activity recognition is an intrinsic component of many pervasive computing and ambient intelligent solutions. This has been facilitated by an explosion of technological developments in the area of wireless sensor network, wearable and mobile computing. Yet, delivering robust activity recognition, which could be deployed at scale in a real world environment, still remains an active research challenge. Much of the existing literature to date has focused on applying machine learning techniques to pre-segmented data collected in controlled laboratory environments. Whilst this approach can provide valuable ground truth information from which to build recognition models, these techniques often do not function well when implemented in near real time applications. This paper presents the application of a multivariate online change detection algorithm to dynamically detect the starting position of windows for the purposes of activity recognition.

Assessing app quality through expert peer review: a case study from the gray matters study.

Health apps focused on inciting behavior change are becoming increasingly popular. Nevertheless, many lack underlying evidence base, scientific credibility and have limited clinical effectiveness. It is therefore important that apps are well-informed, scientifically credible, peer reviewed and evidence based. This paper presents the use of the Mobile App Rating Scale (MARS) to assess the quality of the Grey Matters app, a cross platform app to deliver health education material and track behavior change across multi-domains with the aim of reducing the risk of developing Alzheimer's disease. The Gray Matters app shows promising results following reviews from 5 Expert raters, achieving a mean overall MARS score of 4.45 ± 0.14. Future work will involve undertaking of a detailed content analysis of behavior change apps to identify common themes and features which may lead to the successful facilitation of sustained behavior change.

Technology adoption and prediction tools for everyday technologies aimed at people with dementia.

A wide range of assistive technologies have been developed to support the elderly population with the goal of promoting independent living. The adoption of these technology based solutions is, however, critical to their overarching success. In our previous research we addressed the significance of modelling user adoption to reminding technologies based on a range of physical, environmental and social factors. In our current work we build upon our initial modeling through considering a wider range of computational approaches and identify a reduced set of relevant features that can aid the medical professionals to make an informed choice of whether to recommend the technology or not. The adoption models produced were evaluated on a multi-criterion basis: in terms of prediction performance, robustness and bias in relation to two types of errors. The effects of data imbalance on prediction performance was also considered. With handling the imbalance in the dataset, a 16 feature-subset was evaluated consisting of 173 instances, resulting in the ability to differentiate between adopters and non-adopters with an overall accuracy of 99.42 %.

Doorstep: A doorbell security system for the prevention of doorstep crime.

Safety and security rank highly in the priorities of older people on both an individual and policy level. Older people are commonly targeted as victims of doorstep crime, as they can be perceived as being vulnerable. As a result, this can have a major effect on the victim's health and wellbeing. There have been numerous prevention strategies implemented in an attempt to combat and reduce the number of doorstep crimes. There is, however, little information available detailing the effectiveness of these strategies and how they impact on the fear of crime, particularly with repeat victims. There is therefore clear merit in the creation and piloting of a technology based solution to combat doorstep crime. This paper presents a developed solution to provide increased security for older people within their home.