Convolutional Neural Networks (CNN) and Facial Features based Smart Driver Alertness Detection

Abstract

The development of machine learning assists drivers in the form of unmanned automobiles etc. Nearly 22% of accidents are caused by the driver's drowsiness and lack of alertness. This study intends to monitor the alertness of driver by analyzing their facial expressions and eye variations. The developed smart processing system substantially reduces highway accidents. The proposed model considers various facial characteristics like PERCLOS, Eye Aspect Ratio (EAR), blink values, and yawning. The driver is continually observed by the camera in the proposed method. HC classifiers are used to identify the driver's face. CNN uses the extracted eye images to determine whether the eyes are closed. Based on the classification results, EAR is measured. If the mouth is detected to be open, the lip distance (yawn) is measured. Finally, a warning will be generated if the driver is in a drowsy state.