Preliminary investigation of unconstrained person identification for tabletops using soft biometrics

This paper tries to realize unconstrained person identification for tabletop systems using a ceiling-mounted depth camera that overlooks a table. We extract a user's soft biometrics from a depth image, such as the shoulder width and shape of the head that can be captured from the ceiling. We then try to achieve robust person identification by combining each of the soft biometric. Person identification from a ceiling has several advantages: a ceiling-mounted camera hardly suffers from the occlusion problem, does not interfere users' activities and can capture multiple users in one shot. In the preliminary experiment, proposed method has achieved 83% accuracy identifying 19 participants.