Vehicle Based Driver Drowsiness Detection By Support Vector Machines

Abstract

Driver Drowsiness Remains A Major Cause Of Road Accidents. This Paper Presents A Comparative Analysis Of Indirect Driver Monitoring Systems (DMS) Using Vehicle-based Features, Direct DMS Using Driver-based Facial Behavior, And A Hybrid Approach Combining Both. The System Employs Image Processing And Convolutional Neural Networks (CNNs) To Detect Facial Landmarks Like Eye Aspect Ratio And Mouth Opening, And Classifies Drowsiness States Using Support Vector Machines (SVM). Experimental Validation Using A Dataset From 70 Participants Revealed That The Hybrid DMS Yielded The Highest Balanced Accuracy Of 87.7%, Slightly Outperforming Direct DMS (87.1%) And Significantly Outperforming Indirect DMS (77.9%)

Keywords