Activity Recognition and Dementia Care in Smart Home

Abstract

Smart home is a ubiquitous environment that aims to offer Ambient Assisted Living (AAL) to its occupants. The activity modeling framework proposed in this research work skillfully integrates ambient intelligence into the home environment by a collective process of activity recognition, abnormality detection, and decision making. Moreover, the activity modeling strategy employed in this research work efficiently models both the data and domain knowledge for activity recognition. The primary task in designing an activity recognition system involves the construction of activity model that represents occupant’s Activities of Daily Living (ADL). To achieve activity recognition and abnormality detection competently, it is essential for the activity modeling strategy to consider the design challenges of uncertainty modeling, contextual modeling, composite modeling, activity diversity, and activity dynamics. The challenges of activity dynamics and data uncertainty are well addressed through data-driven approaches, whereas the challenges of activity granularity, contextual knowledge, and activity diversity are well addressed through knowledge-driven approaches. Therefore, activity recognition frameworks are proposed in this research work, where the first framework represents the activity model as a Markov Logic Network and the second framework represents the activity model as a probabilistic ontology. Each of these approaches offers both uncertainty and contextual modeling for activity recognition by integrating data-driven and knowledge-driven techniques. Moreover, this research proposes an assistive dementia care system through smart home that offers functional assistance to the dement occupant during critical situations without the help of caretaker.