Human Fatigue Characterization and Detection Using the Eyelid State and Kalman Filter

One of the most promising commercial applications of Human Computer Interface is the vision based Human fatigue detection systems. Most methods and algorithms currently rely heavily on movement of the head and the colorization of the eye ball. In this paper, a new algorithm for detecting human fatigue by relying primarily on eyelid movements as a facial feature is proposed. The features of the eye region and eyelid movement which are geometric in nature are processed alongside each other to determine the level of fatigue of a person. Haar classifiers are employed to detect the eye region and eyelid features. The eye region is, however processed to ascertain attributes of eyelid movement of each individual of interest. The eyelids are then detected as either opened, closed or in transition state. The movement or velocity of the eyelid is tracked using a Kalman filtered velocity function. This algorithm calculates a human blink cycle for each individual, and estimates the associated errors of the eye movement due to friction using the Kalman filter. The study has established human blink cycle calculation as a new classifier to characterize human fatigue and the calculation of the movement of eyelid using the Kalman filter in determining the level of fatigue.