In this study, five kinds of state variables characterizing the eye blinking behaviour for detecting low wakefulness level of car drivers were investigated. This investigation was carried out in a driving simulator for clarifying what kind of state variables are useful for designing the algorithm of awaking alarm, in terms of minimizing the missing alarm in low wakefulness condition and the excess alarm in awaking condition simultaneously. In applying the ratio regarding opening/closing time of eyelid and the variation of this ratio for designing the algorithm, the time frequency of the missing alarm was approximately 1.5% and extremely lower than the conditions when the algorithm was based on the other state variables like the total time for eye closing or the opening rate of eyelid. The time frequency of excess alarms in applying this algorithm is also not so high. An effectiveness of awaking alarm based on this algorithm was also evaluated and clarified that the awaking effects by the acoustic alarm evaluated in terms of the fluctuation of RRI (ECG) was not sufficient, and the methodology for awaking the driver should be studied.
