Activity recognition in resource-constrained pervasive systems

Abstract: There is an increasing need for personalised and context-aware services in our everyday lives and we rely on mobile and wearable devices to provide such services. Data collected from these devices includes important information about users’ movements, locations, physiological status, and environment. This data can be analysed in order to recognise users’ activities and thus provide contextual information for services. Such activity recognition is an important tool for personalising and adapting assistive services and thereby increasing the usefulness of them.This licentiate thesis focuses on three important aspects for activity recognition usingwearable, resource constrained, devices in pervasive services. Firstly, it is investigated how to perform activity recognition unobtrusively by using a single tri-axial accelerometer. This involves finding the best combination of sensor placement and machine learning algorithm for the activities to be recognized. The best overall placement was found to be on the wrist using the random forest algorithm for detecting Strong-Light, Free-Bound and Sudden-Sustained movement activities belonging to the Laban Effort Framework.Secondly, this thesis proposes a novel machine learning algorithm suitable for resource-constrained devices commonly found in wearable and pervasive systems. The proposed algorithm is computationally inexpensive, parallelizable, has a small memory footprint, and is suitable for implementation in hardware. Due to this, it can reduce battery usage, increase responsiveness, and also make it possible to distribute the machine learning task, which enables balancing computational costs against data traffic costs. The proposed algorithm is shown to have a comparable accuracy to that of more advanced machine learning algorithms mainly for datasets with two classes.Thirdly, activity recognition is applied in a personalised and pervasive service for im-proving health and wellbeing. Two monitoring prototypes and one coaching prototype were proposed for achieving positive behaviour change. The three prototypes were evaluated in a user workshop with 12 users aging between 20 and 60. Participants of the workshop believed that the proposed health and wellbeing app is something people are likely to use on a permanent basis.By applying results from this thesis, systems can be made more energy efficient andless obtrusive while still maintaining a high activity recognition accuracy. It also shows that pervasive and wearable systems using activity recognition have the potential of relieving some problems in health and wellbeing that society face today.