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.
This paper proposes a non-obtrusive system that measures upper extremity velocity during post-stroke rehabilitation exercises within a home environment using a Distance2Go Doppler radar sensing solution. Experimental results are compared with that of an accelerometer in instances of flexion, extension, abduction and adduction. The results demonstrated the added advantages of using a non-obtrusive sensing solution for upper extremity velocity measurement.
Patients with dementia often suffer from stress episodes that escalate to anxiety. This paper presents a feasibility study of using environmental smart microphones to detect anxiety in patients with dementia. It is based on the identified auditory manifestations of anxiety. To have a better understanding of the anxiety manifestations in patients with dementia, 70 diagnosed patients were observed in-situ and 4 caregivers were interviewed. The design of an environmental smart microphone called AnxiCare has been developed and it is introduced. Feasibility interviews regarding the use of AnxiCare were conducted with caregivers at a care residence in Spain. Results from the observations, interviews and a preliminary validation are presented.
Inadvertent falls can cause serious, and potentially fatal injuries, to at risk individuals. One such community of at-risk individuals is the elderly population where age related complications, such as osteoporosis and dementia, can further increase the incidence and negative impact of such falls. Notably, falls within that community has been identified as the leading cause of injury related preventable death, hospitalization and reduction to quality of life. In such cases, rapid detection of, and reaction, to fall events has shown to be critical to reduce the negative effects of falls within this community. Currently, a range of fall detection solutions exist, however, they have several deficiencies related to the core approach that has been adopted. This study has developed an ensemble of thermal vision-based, big data facilitated, solutions which aim to address some of these deficiencies. An evaluation of these logical and data-driven processes has occurred with the promising results presented within this manuscript. Finally, opportunities future work and real-world evaluation have occurred and are underway.
This paper presents a probabilistic approach for the identification of abnormal behaviour in Activities of Daily Living (ADLs) from sensor data collected from 30 participants. The ADLs considered are":" (i) preparing and drinking tea, and (ii) preparing and drinking coffee. Abnormal behaviour identified in the context of these activities can be an indicator of a progressive health problem or the occurrence of a hazardous incident. The approach presented considers the temporal aspect of the sequences of actions that are part of each ADL and that vary between participants. The average and standard deviation for the durations of each action were calculated to define an average time and a range in which a behaviour could be considered as normal for each stage and activity. The Cumulative Distribution Function (CDF) was used to obtain the probabilities of abnormal behaviours related to the early and late completion of activities and stages within an activity. The data analysis show that CDF can provide accurate and reliable results regarding the presence of abnormal behaviour in stages and activities that last over a minute. Finally, this approach could be used to train machine learning algorithms for the abnormal behaviour detection.
Data annotation is the process of segmenting and labelling any type of data (images, audio or text). It is an important task for producing reliable datasets that can be used to train machine learning algorithms for the purpose of Activity Recognition. This paper presents the work in progress towards a semi-automated approach for collecting and annotating audio data from simple sounds that are typically produced at home when people perform daily activities, for example the sound of running water when a tap is open. We propose the use of an app called ISSA (Intelligent System for Sound Annotation) running on smart microphones to facilitate the semi-automated annotation of audible activities. When a sound is produced, the app tries to classify the activity and notifies the user, who can correct the classification and/or provide additional information such as the location of the sound. To illustrate the feasibility of the approach, an initial version of ISSA was implemented to train an audio classifier in a one-bedroom apartment.
As life expectancy has increased, more older adults are experiencing social isolation, which can increase the risk of several negative effects on their health. The use of technology by older adults can support the prevention of social isolation by enabling them to be proactive and motivated to establish human interactions with relatives and friends. This paper explores and presents the results of a survey administered to 74 people from the Centre for Independent Living NI (CILNI) to identify the prevalence of social isolation in Northern Ireland, and the desirability to use technology to prevent it. This paper also introduces an upcoming trial into the use of technology for the prevention of social isolation for older adults living alone.
In this paper we introduce the Worktivity mobile app as a potential solution to help reduce occupational sitting in an office environment. Worktivity functions by sending hourly reminders to stand up or move in addition to showing factual information related to the benefits of being active within the office environment. The Worktivity app was used over a period of 8 weeks by 37 participants aged between 18 and 65 years old from two private office worksites in Northern Ireland. Results demonstrated how users responded to the reminders over the duration of the study with mean acknowledgement rates of 66.06% and 51.57% when using the app and when using the app with a standing desk, respectively.
Purpose":" Occupational sedentary behaviour (SB) is a public health concern associated with negative physical and psychological health consequences (1, 2). Reducing SB in office workers can be challenging due to the need to complete desk-bound work. There is evidence to suggest sit-stand workstations can reduce SB without negative effects on productivity (3), but are costly for companies to implement The aim was to evaluate the effects of a mobile app-based intervention targeting reductions in occupational SB, delivered with and without sit-stand workstations, on employee mood and productivity. Methods This 8-week feasibility cluster randomised controlled trial recruited desk-based office workers (n=56), aged 18-65 years, from three worksites in Northern Ireland. Following baseline measures, worksites were randomised to one of three groups":" mobile app; mobile app and sit-stand workstation; or control. The "Worktivity" app, developed using the Behaviour Change Wheel (4) encouraged office workers to reduce sitting by self-monitoring SB and setting 'sit-less' goals. The app also delivered 'sit-less' nudges, educational prompts and progress reports. Mood and work productivity were measured at baseline, four and eight weeks. Productivity was measured daily for five consecutive days using ecological momentary assessment via text-message/e-mail where participants responded to a question relating to work productivity. Mood was measured using the Brunel Mood Scale (BRUMS). Results The intervention will conclude in December 2017 and findings will be available for ISBNPA 2018. As this is a feasibility trial, analysis will be mainly descriptive. Investigations will be exploratory to provide estimates of key parameters and inform the design of a definitive trial. Conclusions This study represents the first investigation of the effects of an app intervention, designed to reduce occupational SB, on employee productivity and mood. Findings are expected to inform the development of a larger-scale m-health intervention to reduce SB in office workers.
This paper discusses the opportunities and challenges associated with the collection of a large scale, diverse dataset for Activity Recognition. The dataset was collected by 141 undergraduate students, in a controlled environment. Students collected triaxial accelerometer data from a wearable accelerometer whilst each carrying out 3 of the 18 investigated activities, categorized into 6 scenarios of daily living. This data was subsequently labelled, anonymized and uploaded to a shared repository. This paper presents an analysis of data quality, through outlier detection and assesses the suitability of the dataset for the creation and validation of Activity Recognition models. This is achieved through the application of a range of common data driven machine learning approaches. Finally, the paper describes challenges identified during the data collection process and discusses how these could be addressed. Issues surrounding data quality, in particular, identifying and addressing poor calibration of the data were identified. Results highlight the potential of harnessing these diverse data for Activity Recognition. Based on a comparison of six classification approaches, a Random Forest provided the best classification (F-measure":" 0.88). In future data collection cycles, participants will be encouraged to collect a set of “common” activities, to support generation of a larger homogeneous dataset. Future work will seek to refine the methodology further and to evaluate model on new unseen data.