A knowledge-based approach towards human activity recognition in smart environments

For many years it is known that the population of older persons is on the rise. A recent report estimates that globally, the share of the population aged 65 years or over is expected to increase from 9.3 percent in 2020 to around 16.0 percent in 2050 [1]. This point has been one of the main sources of motivation for active research in the domain of human activity recognition in smart-homes. The ability to perform ADL without assistance from other people can be considered as a reference for the estimation of the independent living level of the older person. Conventionally, this has been assessed by health-care domain experts via a qualitative evaluation of the ADL. Since this evaluation is qualitative, it can vary based on the person being monitored and the caregiver’s experience. A significant amount of research work is implicitly or explicitly aimed at augmenting the health-care domain expert’s qualitative evaluation with quantitative data or knowledge obtained from HAR. From a medical perspective, there is a lack of evidence about the technology readiness level of smart home architectures supporting older persons by recognizing ADL [2]. We hypothesize that this may be due to a lack of effective collaboration between smart-home researchers/developers and health-care domain experts, especially when considering HAR. We foresee an increase in HAR systems being developed in close collaboration with caregivers and geriatricians to support their qualitative evaluation of ADL with explainable quantitative outcomes of the HAR systems. This has been a motivation for the work in this thesis. The recognition of human activities – in particular ADL – may not only be limited to support the health and well-being of older people. It can be relevant to home users in general. For instance, HAR could support digital assistants or companion robots to provide contextually relevant and proactive support to the home users, whether young adults or old. This has also been a motivation for the work in this thesis. Given our motivations, namely, (i) facilitation of iterative development and ease in collaboration between HAR system researchers/developers and health-care domain experts in ADL, and (ii) robust HAR that can support digital assistants or companion robots. There is a need for the development of a HAR framework that at its core is modular and flexible to facilitate an iterative development process [3], which is an integral part of collaborative work that involves develop-test-improve phases. At the same time, the framework should be intelligible for the sake of enriched collaboration with health-care domain experts. Furthermore, it should be scalable, online, and accurate for having robust HAR, which can enable many smart-home applications. The goal of this thesis is to design and evaluate such a framework. This thesis contributes to the domain of HAR in smart-homes. Particularly the contribution can be divided into three parts. The first contribution is Arianna+, a framework to develop networks of ontologies - for knowledge representation and reasoning - that enables smart homes to perform human activity recognition online. The second contribution is OWLOOP, an API that supports the development of HAR system architectures based on Arianna+. It enables the usage of Ontology Web Language (OWL) by the means of Object-Oriented Programming (OOP). The third contribution is the evaluation and exploitation of Arianna+ using OWLOOP API. The exploitation of Arianna+ using OWLOOP API has resulted in four HAR system implementations. The evaluations and results of these HAR systems emphasize the novelty of Arianna+.