• First study to identify physiological correlates of social disconnection in the real-world. • Detected mood-related general social connectedness using wearable devices. • Developed a social density estimation model of people within a 5m vicinity of participants. • Detected social connectedness to people nearby. • Classification models detected correlates of social connection using physiological features from wearable devices. Loneliness, low mood, and social disconnection can have damaging effects on physical and mental health. Detecting these emotions in the context of everyday life is important as these psychological states can manifest differently outside the laboratory. Lifelogging and quantified self technologies, including wearable devices, offer an approach to continuously monitor physiological signals in the real-world. However, little work has been undertaken with these devices to detect social connection in everyday life. This paper presents a study that leveraged machine learning to infer mood and social connectedness in everyday life using multimodal lifelogging data collected via a wrist-worn wearable device. Fifty participants were supplied with a wearable device and smartphone that collected physiological and subjective data across two consecutive weekdays as they went about their daily lives. The analysis examined physiological correlates between a person’s psychological perceptions of general connectedness, as well as feelings of in-the-moment social connection to people within an estimated 5m vicinity and mood at specific timepoints using four machine learning classification models – k-Nearest Neighbour, Random Forest, Support Vector Machine and Naïve Bayes. Results demonstrated that Random Forest obtained the highest accuracy of 0.8 – 0.84 for the binary detection of mood and social connection in everyday life.

Advancements in wearable technologies have made the use of physiological signals, such as Electrodermal Activity (EDA) and Heart Rate Variability (HRV), more prevalent for detecting changes in the autonomic nervous system within virtual reality (VR). However, the challenge lies in utilizing these signals to objectively detect presence in VR, which typically relies on self-reports that can be inherently biased. This paper addresses this issue and presents a study ( N =26) that investigates the effect that different levels of presence has on physiological responses in VR. A neutral VR environment was created that incorporated three levels of presence (high, medium and low) that were invoked by tuning different parameters. Participants wore a wrist-worn wearable device that captured their physiological signals whilst they experienced each of these environments. Results indicated that tonic and phasic components of the EDA signal were significant in differentiating between the levels. Two novel features, constructed using both the phasic and tonic components of EDA, successfully differentiated between presence levels. Analysis of the HRV data illustrated a significant difference between the low and medium levels using the ratio between low frequency to high frequency. • Validation of the design of three different levels of presence in a VR environment. • Development of two novel features (cf1, cf2) significantly differentiated between levels of presence. • Poincare maps illustrate the variability in the data between different presence levels.

Training systems are used in many industries, ranging from surgery to space missions to rehabilitation. Virtual Reality (VR) is a technology that has been incorporated as an effective tool in such training systems to simulate the environment, especially in situations where the training can’t take place in the actual environment. For a training environment and task to be effective, it must sufficiently challenge the trainee. One parameter that can be used to measure this is cognitive load (CL), which is defined as the amount of working memory used while performing a learning task. This parameter needs to be sufficiently high to maximize learning but not too high as to overload the trainee. However, the challenge is to detect this state using objective physiological measures, which can be collected during the entire task. This paper presents a study to classify CL using a combination of Electroencephalogram (EEG) and Electrodermal Activity (EDA) signals during a procedural VR training task. Thirty participants undertook a study where they built a designated model within a given time over multiple levels that were constructed to induce low to high CL. Features generated from the data were subject to feature selection (FS), which was undertaken using the Mutual Information (MI) technique. Binary classification models were developed using Support Vector Machines (SVM), Random Forest (RF), k-Nearest Neighbors (kNN), Extreme Gradient Boosting (Xgboost) and Multi-Layer Perceptrons (MLP). Results illustrated that the Xgboost classifier performed the best with an F1-score of $0.831 \pm 0.030$ and accuracy of $0.805 \pm 0.033.$ SHAP analysis of the features illustrated greater contributions from the frontal and occipital regions of the brain and frequency domain features from tonic skin conductance.

Abstract Prolonged and repetitive stress on muscles, tendons, ligaments, and nerves can have long-term adverse effects on the human body. This can be exasperated while working if the environment and nature of the tasks puts significant strain on the body, which may lead to work-related musculoskeletal disorders (WMSDs). Workers with WMSDs can experience generalized pain, loss of muscle strength, and loss of ability to continue working. Most WMSDs injuries are caused by ergonomic risks, such as repetitive physical movements, awkward postures, inadequate recovery time, and muscular stress. Fatigue can be seen as a detector of ergonomic risk, as the accumulation of fatigue can significantly increase the possibility of injury. Thirty participants completed a series of repetitive physical tasks over a six-hour period while wearing sensors to capture data related to heart rate and movement, while external embedded sensors captured ground reaction and hand exertion force. They also provided subjective ratings of fatigue at the start and end of the experiment. Classifiers for fatigue (high vs low) were constructed using three methods: linear discriminant analysis (LDA), k-nearest neighbor (kNN), and polynomial kernel-based SVM (P-SVM) and were validated using a tenfold cross-validation technique that was repeated a hundred times. Results of our supervised machine learning approach demonstrated a maximum accuracy of 94.15% using P-SVM for the binary classification of fatigue.

Presence, the feeling of being there, is an important factor that affects the overall experience of Virtual Reality (VR). Higher presence commonly provides a better experience in VR than lower presence. However, presence is commonly measured subjectively through postexperience questionnaires, which can suffer from participant biases, dishonest answers, and fatigue. It can also be difficult for subjects to accurately remember their feelings of presence after they have left the VR experience. In this paper, we measured the effects of different levels of presence (high and low) in VR using physiological and neurological signals. The experiment involved 24 participants in a between-subjects design. Results indicated a significant effect of presence on both physiological and neurological signals. We noticed that higher presence results in higher heart rate, less visual stress, higher theta and beta activities in the frontal region, and higher alpha activities in the parietal region. These findings and insights could lead to an alternative objective measure of presence.

Stress is a negative emotion that is part of everyday life. However, frequent episodes or prolonged periods of stress can be detrimental to long-term health. Nevertheless, developing self-awareness is an important aspect of fostering effective ways to self-regulate these experiences. Mobile lifelogging systems provide an ideal platform to support self-regulation of stress by raising awareness of negative emotional states via continuous recording of psychophysiological and behavioral data. However, obtaining meaningful information from large volumes of raw data represents a significant challenge because these data must be accurately quantified and processed before stress can be detected. This work describes a set of algorithms designed to process multiple streams of lifelogging data for stress detection in the context of real world driving. Two data collection exercises have been performed where multimodal data, including raw cardiovascular activity and driving information, were collected from 21 people during daily commuter journeys. Our approach enabled us to 1) pre-process raw physiological data to calculate valid measures of heart rate variability, a significant marker of stress, 2) identify/correct artefacts in the raw physiological data, and 3) provide a comparison between several classifiers for detecting stress. Results were positive and ensemble classification models provided a maximum accuracy of 86.9 percent for binary detection of stress in the real-world.

This work introduces a set of scalable algorithms to identify patterns of human daily behaviors. These patterns are extracted from multivariate temporal data that have been collected from smartphones. We have exploited sensors that are available on these devices, and have identified frequent behavioral patterns with a temporal granularity, which has been inspired by the way individuals segment time into events. These patterns are helpful to both end-users and third parties who provide services based on this information. We have demonstrated our approach on two real-world datasets and showed that our pattern identification algorithms are scalable. This scalability makes analysis on resource constrained and small devices such as smartwatches feasible. Traditional data analysis systems are usually operated in a remote system outside the device. This is largely due to the lack of scalability originating from software and hardware restrictions of mobile/wearable devices. By analyzing the data on the device, the user has the control over the data, i.e., privacy, and the network costs will also be removed.

The World Wide Web (WWW) has radically changed the way in which we access, generate and disseminate information. Its presence is felt daily and with more internet-enabled devices being connected the web of knowledge is growing. We are now moving into era where the WWW is capable of `understanding' the actual/intended meaning of our content. This is being achieved by creating links between distributed data sources using the Resource Description Framework (RDF). In order to find information in this web of interconnected sources, complex query languages are often employed, e.g. SPARQL. However, this approach is limited as exact query matches are often required. In order to overcome this challenge, this paper presents a probabilistic approach to searching RDF documents. The developed algorithm converts RDF data into a matrix of features and treats searching as a machine learning problem. Using a number of artificial neural network algorithms, a successfully developed prototype has been developed that demonstrates the applicability of the approach. The results illustrate that the Voted Perceptron classifier (VPC), perceptron linear classifier (PERLC) and random neural network classifier (RNNC) performed particularly well, with accuracies of 100%, 98% and 93% respectively.

Worldwide the rate of preterm birth is increasing, which presents significant health, developmental and economic problems. Current methods for predicting preterm births at an early stage are inadequate. Yet, there has been increasing evidence that the analysis of uterine electrical signals, from the abdominal surface, could provide an independent and easy way to diagnose true labour and predict preterm delivery. This analysis provides a heavy focus on the use of advanced machine learning techniques and Electrohysterography (EHG) signal processing. Most EHG studies have focused on true labour detection, in the window of around seven days before labour. However, this paper focuses on using such EHG signals to detect preterm births. In achieving this, the study uses an open dataset containing 262 records for women who delivered at term and 38 who delivered prematurely. The synthetic minority over sampling technique is utilized to overcome the issue with imbalanced datasets to produce a dataset containing 262 term records and 262 preterm records. Six different artificial neural networks were used to detect term and preterm records. The results show that the best performing classifier was the LMNC with 96% sensitivity, 92% specificity, 95% AUC and 6% mean error.

There is growing global concern over the growing levels of obesity and the fact that people in general are not as active as they once were. Many believe that this is directly related to poor diet and our increasing reliance on technology, such as television, social networking, computer games, and voice activated home control systems. These kinds of activities increase sedentary behaviour across all age groups and is considered one of the main contributors to obesity and poor health. For this reason decreasing sedentary behaviour is considered a crucial theme within many research programs in health. Ironically, there is general agreement that the use of technology is likely to help researchers understand this type of behaviour. One interesting approach is based upon the use of human digital memories to provide visual lifelogs of a user's activity and to identify the behaviour patterns of individuals. In this way visual lifelogs provide a way for user's to evaluate their lifestyle choices. This paper discusses some of the key technologies used to achieve this and considers some of the challenges that still need to be addressed.

There is growing concern, among senior personnel in the National Health Service in the UK, over the increased development of pressure ulcers. The occurrence of pressure ulcers has been attributed to prolong sedentary behaviour. Providing care, for this preventable condition, is costly and time-consuming for patients and medical practitioners. Extra bedside assistance is needed; however, with the workload of medical staff increasing, this is not always practical. In order to prevent the occurrence of pressure ulcers new and novel ways of remotely monitoring patients is essential. An interesting approach worth considering is the use of human digital memories, which provide visual life logs of a patient's physiological and environmental data. This paper discusses some of the current technologies used within the area and how they might be applied to the management and prevention of pressure ulcers. We have successfully developed a working prototype system to demonstrate the applicability of our approach.