Individualized drowsiness detection during driving by pulse wave analysis with neural network

This paper presents a detection method of driver's drowsiness with focus on analyzing individual differences in biological signals and performance data. We have studied biological signals of a driver to detect drowsiness during driving. Our former research suggested a method analyzing changes in indexes derived from biological signals, however the method needs to be configured for each driver because the relation between the indexes and the drowsiness depends on individuals. To analyze the indexes in consideration of the individual differences, neural networks was used in this paper. The learning function the networks was utilized to adapt to the differences. We conducted a experiment that 6 drivers drove a driving simulator to gather their pulse wave and steering data. As the result of learning and analyzing the indexes in neural networks, 98% of the highest ratio was shown in detection of driver's drowsiness. A method of detecting driver's drowsiness is a need for realization of safer traffic environment. The proposed method would contribute to prevent traffic accidents caused by human errors in a drowse.