Gaze detection from different view points of cameras

In order to build "intelligent rooms", it is necessary to establish a system to classify the subject's focus of attention, which might be focused on various surrounding objects. In this paper, we propose a system to detect eye contact with targets attached to each camera positioned in a room. Each camera detects faces and estimates the position of the head. Our system only detects eye regions and irises from a near frontal view of the face. Assuming that the shape of the iris denotes a real circle at direct eye contact, our system evaluates the roundness of the circle by values in the parameter space with the circular Hough Transform, and determines the approximate gaze direction.