Assessment of Drowsy Drivers by Fuzzy Logic approach based on Multinomial Logistic Regression Analysis

摘要

The aim of this study is to investigate the effect of behavioral and physiological measures for predicting driver’s drowsiness in order to develop an intelligent transportation system such as fuzzy logic for preventing fatal traffic accidents by evaluating the lack of driver’s arousal level. In this paper, behavioral and physiological measures are considered but because of high costs of measuring physiological measures in laboratories, only behavioral measures are examined. Drowsy states of drivers were predicted by means of the multinomial logistic regression model which are independent variables and a dependent variable in order behavioral measures and driver’s drowsiness, respectively. For better understanding of the multinomial logistic regression model related to drowsy states, all behavioral measures were entered into the model. It was found that behavioral measures were investigated with a significant coefficient of 0.05 according to statistical science which is ANOVA. From results of statistical view, prediction accuracy and probability of behavioral measures, it is clear that the most predicted behavioral measure is Neck bending angle (vertical) with regression coefficient (R2) 0.74, correlation coefficient 0.56, with probability of 0.78, and average prediction accuracy amongst drowsy groups 0.73, which represents a good fitness in the model. Furthermore, driver’s sleep behavior in travel distances and weather conditions was simulated in fuzzy logic for understanding the effect of these conditions over driver’s sleep behavior. Finally, Fuzzy logic showed that driver’s sleep behavior in unsuitable weather such as rainy condition is affected in high risk of drivers’s drowsiness level in comparison with light condition that drivers have lower drowsiness.