Driver Inattention Monitoring System Based on the Orientation of the Face Using Convolutional Neural Network

Driving inattentively is one of the prime reasons for vehicle accidents worldwide and has significant implications for road safety. A prompt alert to the inattentive driver can mitigate many accidents and save numerous lives, and reduce the cost of damages caused by accidents. To achieve this, a proposal of a nonintrusive and noninvasive driver inattention monitoring and alerting system in real time has been put forward. A mobile camera mounted on the windshield captures the video of the driver. Viola-Jones algorithm detects the face in each frame of the video and the Kanade-Lucas-Tomasi (KLT) algorithm tracks the detected face from one frame to another frame. The driver is classified as inattentive or attentive using Convolutional Neural Network (CNN). The transfer learning of the AlexNet Convolutional Neural Network architecture is adopted for the classification. The accuracy, precision, sensitivity, F1 score, and specificity of the system proposed in this paper are 98.24%, 100%, 96.47%, 98.21% and 100%, respectively.

I. Introduction

Inattentive driving is one of the prime reasons for vehicle accidents worldwide. The National Highway Traffic Safety Administration (NHTSA) reports 2840 fatalities and 4 lakhs injured people due to distracted driving in 2018 [1]. According to the National Crimes Record Bureau (NCRB), major causes of road accident deaths are dangerous or careless driving (27.3%) [2]. Driver inattention indicates reduced attention to critical activities for safe driving. An inattentive driver may lose his ability to control the vehicle, and might suddenly deviate from the road and hit another vehicle, pedestrian or other obstacles. The various factors contributing to this are drowsiness and distractions.