Study on Fatigue Driving Detection Model Based on Steering Operation Features and Eye Movement Features

Fatigue driving is one of the invisible killers on the expressway, which can lead to serious traffic accidents. Therefore, the development of real-time driver fatigue detection and warning devices can effectively reduce the possibility of road accidents caused by fatigue driving. Based on the theoretical research of fatigue driving and information fusion, fatigue driving experiment was carried out using driving stimulation platform and eye tracker to record steering operation data and eye movement data of 10 drivers. Total of 25 fatigue indicators were extracted, 23 of which were selected with valid statistical significance. 9-dimensional optimal subset was filtered out from these 23 indicators by sequential floating forward selection method and support vector machine classification. Finally, real-time fatigue detection model was built based on sliding time window principle with an accuracy of 91%, sensitivity of 94% and specificity of 88%.