Detecting of Fatigue States of a Car Driver

This paper deals with research on fatigue states of car drivers on freeways and similar roads. All experiments are performed on-the-road. The approach is based on the assumption that fatigue indicators can be derived from driver+car system behaviour by measuring and processing appropriate factors. For our experiments we designed an array of devices to measure selected physiological and technical parameters. On the basis of experiments already performed and described in the literature, we selected signals that carry appropriate information about fatigue states of a driver. Two data sets are compared: the first set was obtained from alert drivers, and the second one from provably sleep-deprived drivers. We are trying to calibrate “fatigue states” of the driver through trained rater estimates, and to extract reliable symptoms from the measured signals; we are seeking relations between the symptoms and raters fatigue estimates. The results of experiments already performed indicate that the values of symptoms evolve in cycles, which is essential to take into account during design of classifiers in the future.