A Novel Infrared Motion Sensing System for Compressive Classification of Physical Activity

Infrared radiation changes (IRCs) induced by human activity can provide important information about activity patterns. This paper presents an IRC-based compressive classification method for recognizing the physical activities of interest in home-based assisted living. To fully capture the IRC compressively, a multi-view infrared motion sensing system is developed, which consists of three IRC sensing modules, that is, one module on the ceiling and two modules on opposite tripods facing each other. A pilot study is conducted in the context of classifying six typical physical activities with the incorporation of the classification techniques, including hidden Markov model and support vector machine, which demonstrates the effectiveness of our system.