A live video analytic system for affect analysis in public space

We propose a video analytic system capable of analyzing group density and social signals such as affect in public space. In contrast to most facial analysis systems that operate on fixed cameras, this demonstration incorporates Pan-Tilt-Zoom (PTZ) camera control and facial video analysis hand in hand, in order to effectively locate individuals, distill expression signals and provide visual feedback in realtime. We assume individuals can sit on a fixed number of seats or benches. Our PTZ control module operates cooperatively with a face analysis module to actively search for facial shots in the public space and perform expression and pose analysis. Group density is estimated as the occupancy rate with respect to the possible seating locations and group-level affect is inferred from aggregate facial expression recognition.