Towards predicting driver intentions from patterns of eye fixations

It is generally believed that the pattern of eye fixations reflects, to some degree, the cognitive state of the observer. Studies have shown that eye movement patterns can be correlated to certain mental states in object recognition or to particular driving situations, such as following another vehicle. In this paper, the author explores the possibility of using these correlations in an intelligent vehicle system to ascertain whether the driver is performing certain actions. Such knowledge is necessary for seamlessly interfacing active control systems with the driver as well as managing the display of in-vehicle information or warnings. Experiments with a system based on hidden Markov dynamic models, which are related to Markovian analysis used in eye movement studies, have produced promising results. He discusses one advantage of incorporating eye movements into such a system as well as how it might be accomplished. (A) For the covering abstract see ITRD E106152.