AAAI Publications, Workshops at the Twenty-Eighth AAAI Conference on Artificial Intelligence

Font Size: 
Activity Prediction Based on Tme Series Forcasting
Mohamed Tarik Moutacalli, Kevin Bouchard, Abdenour Bouzouane, Bruno Bouchard

Last modified: 2014-06-18

Abstract


Activity recognition is a crucial step in automaticassistance for elderly and disabled people, such asAlzheimer’s patients. The large number of activities ofdaily living (ADLs) that these persons are used to per-forming as well as their inability, sometimes, to start anactivity make the recognition process difficult, if not im-possible. To adress such problems, we propose a time-based activity prediction approch as a preliminary stepto activity recognition. Not only it will facilitate therecognition, but it will also rank activities according totheir occurrence probabilities at every time interval. Inthis paper, after detecting activities models, we imple-ment and validate an activity prediction process using atime series framework.

Keywords


Smart Homes; Time series; Activity prediction; activity recognition

Full Text: PDF