Non-rigid eye movement tracking and eye state quantification

Tracking and quantifying the eye's state in real-time whilst also accounting for object size variation is crucial in computer vision systems for Human Computer Interaction (HCI) applications and fatigue detection techniques. This paper presents a novel approach to track and quantify the eye states using the Mean Shift tracking algorithm enhanced by an adaptive scale. The iris is used as a target model for Mean Shift tracking and the moment features are used to estimate the iris area in order to quantify the state of the eye. The proposed approach to tracking non-rigid eye movement and to providing eye state quantification is shown to produce very accurate results.