Abstract
By monitoring both driving behavior and eye condition and by using them as input of the drowsiness evaluation algorithm, is possible to determine whether the driver is in a state of fatigue.
The first kind of input is given by the driver visual behavior: PERCLOS ( PERcentage of eye CLOSure test) [1], blink frequency, eyelid motion. For this purpose a real time acquisition of driver image is done by means of an active infrared illuminator, each of the above mentioned visual parameters is correlated with reference patterns related to driver state with and without drowsiness.
The second kind of input is the driving behavior , taken into account by measuring the accelerator pedal activity, the measured pattern is used to correlate in real time by means of a mathematical model with reference patterns related to drivers with and without drowsiness, in particular, the decay of the pedal force pattern has been identified as an important information.
Finally both kinds of input are combined by means of a fuzzy logic classifier to infer the level of drowsiness of the driver. Based on the results, audio and visual warnings[2] are given to the driver.
KEYWORDS – fatigue driving, PERCLOS ,driving behavior, fuzzy logic classification, warning.