An efficient Real-Time Driver Drowsiness Monitoring System by Using Ensembled Regression Trees

Abstract

Driver drowsiness, primarily resulting from fatigue and insufficient sleep, is a leading contributor to road accidents worldwide. Data analytics reveals that driver drowsiness is the reason for one-fifth of all traffic accidents worldwide. Early detection of drowsiness is vital to prevent accidents. Many methods have been developed for early detection of drowsiness and alerting sleepy drivers. Convolutional Neural Networks (CNNs) are highly effective for detecting drowsiness. However, they are not ideal for real-time applications due to the time-intensive nature of data collection and processing. Additionally, they often suffer from low real-time performance and lack an integrated alerting mechanism. To overcome this drawback, facial landmark detection method is proposed using ensembled regression trees. In the proposed method images are directly captured from the camera and applied 68 facial landmarks images and compared against input image for calculating Eye Aspect Ratio, blink rate, eye closure time, and lane deviation to detect signs of drowsiness whether the driver is drowsy or not. When the driver is drowsy, it plays the alerting alarm. According to experimental analysis, the proposed method ensures reliable accuracy while maintaining low computational latency, making it an efficient and practical solution for real-time driver monitoring compared to traditional deep learning models.