A Novel Driver Fatigue and Distraction Detection Method Based on Improved Swin Transformer

Driving fatigue and distraction are the main causes of traffic accidents. In order to reduce the occurrence of traffic accidents and improve driving safety, with considering the individual differences in driver fatigue and distraction, an improved SE-Swin Transformer network that can simultaneously detect driver fatigue and distraction is developed. Yawning is taken as the main feature of fatigue detection. In order to tackle the individual differences in yawning, the individual-specific mouth aspect ration (MAR) is imployed in the anlaysis of the YawnDD dataset. For distraction detection, the StateFarm dataset is selected, and the different individual driving habits is addressed by training and testing the individual dataset obtained from SFDD dataset. The experimental results show that the integrated detection method established in the paper can achieve 96.68% and 97.23% accuracy for fatigue and distraction respectively, and the accuracy of 96.51% for fatigue and distraction detection simultaneously. The results also show that considering individual differences can improve the accuracy and generalization of the methods.