Nonintrusive Occupant Identification By Sensing Body Shape And Movement

The ability to identify people has numerous applications including in smart buildings where the building can be customized to the needs of its occupants or for other applications such as in assisted living and customer behavior analysis in commercial settings. There are different methods used for occupant identification. Some are intrusive such as using cameras or microphone and others require the users to carry mobile gadgets to be identified. In this paper, we present a nonintrusive method to identify people by sensing their body shape and movement. Such information is derived from using ultrasonic sensors to measure the height and width as the occupant walks through the instrumental doorway. In fact, height and width are not unique to every occupant, but extracting a set of features from the variations in height and width makes identification possible. In this study, our system senses a stream of height and width data, recognizes the walking event when a person walks through the door, extracts features that capture a person's movement as well as physical shape. These features are fed to our clustering algorithm that associates each occupant with a distinct cluster. We deployed our system for 1 month. We found out that our approach achieves 95% accuracy with 20 occupants suggesting the suitability of our approach in commercial building settings. In addition, we found out that using girth to distinguish between occupants is more successful than using height.