Classifying Activity Patterns using Activity Graphs from Sensors

Advancements in sensor technologies give us the opportunity to recognize activities of daily living. Activity pattern classification is a technique that predicts class labels of people based on activity patterns. Because people have both intrinsic and common activity patterns, mining discriminative features reflecting the intrinsic patterns is important for this classification problem. In this paper, we propose an effective method for classifying activity patterns. In order to mine discriminative features, we represent activities as a graph model and mine activity patterns in various periods. Experiments show the proposed method achieves high classification accuracy compared with existing graph classification techniques.