A novel non-intrusive eye gaze estimation using cross-ratio under large head motion

Abstract

Eye gaze estimation systems calculate the direction of human eye gaze. Numerous accurate eye gaze estimation systems considering a user’s head movement have been reported. Although the systems allow large head motion, they require multiple devices and complicate computation in order to obtain the geometrical positions of an eye, cameras, and a monitor. The light-reflection-based method proposed in this paper does not require any knowledge of their positions, so the system utilizing the proposed method is lighter and easier to use than the conventional systems. To estimate where the user looks allowing ample head movement, we utilize an invariant value (cross-ratio) of a projective space. Also, a robust feature detection using an ellipse-specific active contour is suggested in order to find features exactly. Our proposed feature detection and estimation method are simple and fast, and shows accurate results under large head motion.