Contents I. Introduction
A detection method of driver's drowsiness is described in this paper. We have researched driver's biological signals and performance data to detect the drowsiness for the prevention of traffic accidents caused by human errors. Our former research suggested a method to detect driver's drowsiness by observing changes of biological signals, however, the method needs to be configured on each driver because the relation between the data and the drowsiness depends on individuals. To analyze the data in consideration of the individual differences, neural networks was used. Neural networks have effective adaptability for the difference with its learning function. As the input data to the networks, Sympathetic Nerve Activity, Parasympathetic Nerve Activity, Pulse Rate, Lyapunov Exponent, and steering instability, were derived from driver's pulse wave and steering data. Additionally, the score of Epworth Sleepiness Scale (ESS) that is the questionnaire used to determine the level of daytime sleepiness, was also input to the networks. Drivers were individualized by learning these driving features in the networks. We conducted a experiment that 6 drivers drove a driving simulator to evaluate the method. Their pulse wave and steering data were measured while they drove the simulator. As the result of learning and analyzing the features, up to 98 % accuracy was given in detection of the drowsiness. The result was also compared with our previous method at the last. The general outline of this paper is shown in Fig. 1.
