Activity Prediction Based on Time Series Forcasting

Activity recognition is a crucial step in automatic assistance for elderly and disabled people, such as Alzheimer’s patients. The large number of activities of daily living (ADLs) that these persons are used to performing as well as their inability, sometimes, to start an activity make the recognition process difficult, if not impossible. To adress such problems, we propose a timebased activity prediction approch as a preliminary step to activity recognition. Not only it will facilitate the recognition, but it will also rank activities according to their occurrence probabilities at every time interval. In this paper, after detecting activities models, we implement and validate an activity prediction process using a time series framework.