LAUSR

Abstract The human eye and its characteristics (texture, gaze, and movement) have shown to be reliable in expressing human emotional states. A non-invasive eye tracking and state (open/close) estimation system is, consequently, fundamental for developing human-machine interaction systems, hands-off process control, and fatigue detection. In the current work, a novel and fast method based on an optical correlator is proposed for eye detection first, followed by eye state estimation. The second aim of this work was to investigate the simulated optical correlator on a real world driving videos for driver drowsiness detection. For the first time, a numerical simulation of the Vander Lugt Correlator (VLC) was used for automatically detecting the eye center. Further, correlation filters tailored to all the eye states were proposed and their design, to recognize eyes in cluttering and noisy environments, was investigated. Besides, the correlation peak amplitude returned by the VLC based algorithm was investigated to detect the driver drowsiness. Results reported on four international databases reveal that the proposed method improves over the recently published methods in eye tracking. Evaluations of the simulated VLC on real world driving videos from the SHRP2 database demonstrate its good performance on the driver's eye tracking and closure detection.