An Efficient Approach for Appearance Based Eye Gaze Estimation with 13 Directional Points

This paper proposes a real-time eye gaze tracking interface based on an active appearance method using a simple web or smartphone camera in an unconstrained environment, where natural head movements have been taken into account. Here, feature extraction from an eye gaze image has been performed by separating the sclera pixel area from the masked eye image. The performance evaluation has been accomplished using decision tree, random forest, and extra tree classifiers. Over 13 targets, the proposed approach has an approximate 98% accuracy with extra tree classifier. The proposed approach may be beneficial to different eye gazed tracking applications in the field of human-computer interaction, robotics, and medical science.