User indoor localisation system enhances activity recognition: A proof of concept

Older people would like to live independently in their home as long as possible. They want to reduce the risk of domestic accidents because of polypharmacy, physical weakness and other mental illnesses, which could increase the risks of domestic accidents (i.e. a fall). Changes in the behaviour of healthy older people could be correlated with cognitive disorders; consequently, early intervention could delay the deterioration of the disease. Over the last few years, activity recognition systems have been developed to support the management of senior citizensâ\u80\u99 daily life. In this context, this paper aims to go beyond the state-of-the-art presenting a proof of concept where information on body movement, vital signs and userâ\u80\u99s indoor locations are aggregated to improve the activity recognition task. The presented system has been tested in a realistic environment with three users in order to assess the feasibility of the proposed method. These results encouraged the use of this approach in activity recognition applications; indeed, the overall accuracy values, amongst others, are satisfactory increased (+2.67% DT, +7.39% SVM, +147.37% NN)