A Systematic Review on Drivers Drowsiness Detection using Machine Learning Awake Behind the Wheel

Abstract Avoiding accidents on the road and improving transport safety strongly requires proper tracking of driver alertness. In this review, the author has focused on the recent developments of drowsiness detection techniques, which have two main methods: facial procedure metrics and deep learning models. The review focuses on studies with drowsiness detection using EAR (Eye Aspect Ratio), MAR (Mouth Aspect Ratio), and NLR (Nose Length Ratio) which measure eye closure, yawning, and head nodding motion cunningly associated with drowsiness. It examines works done with Convolutional Neural Networks (CNNs) and Long Short-Term Memory (LRCN) that were trained on rich databases with diverse head pose and pupil property features for predicting drowsiness. It analyzes models that used deep learning convolutional neural networks (CNNs) and LRCN for temporal sequences (Long Short-term Memory Cells) to demonstrate a wide range of head poses and pupil attributes for drowsiness prediction. The analytical work shows a combination of drowsiness metric evaluation using real time to enhance deep learning model accuracy achieves better performance when measuring prediction accuracy compared to systems that do not combine metric evaluation with deep analysis. Such models are able to accurately notice drowsiness in a driver and warn them with enough time before accident. With this paper, we intend to give some idea for the research with the advancing technology for driver drowsiness detection to make road transport safer.