A Mobile Crowdsourcing Approach for Smart Edge-Based Driver Drowsiness Detection

Drowsy driving is a major contributor to road accidents, accounting for approximately 15.5% of fatal crashes. With the increasing prevalence of mobile devices and roadside infrastructure, implementing an effective drowsiness detection system can significantly enhance road safety. While numerous approaches have been proposed, most existing solutions lack a distributed framework that ensures efficiency while safeguarding driver privacy. This paper introduces a two-stage Smart Edge-based Driver Drowsiness Detection System that leverages edge computing for real-time analysis. The system utilizes mobile devices within vehicles to monitor driver behavior without transmitting sensitive data. The decision-making process is carried out at the edge, where drowsiness is confirmed by correlating driver condition data from mobile clients with vehicle movement patterns. Our method incorporates: (1) a distributed edge framework with hierarchical nodes—Main Edge Node (MEN) and Local Edge Node (LEN)—for improved data processing, and (2) an optimized data fusion strategy, integrating (i) local detection of drowsiness through facial analysis using a Convolutional Neural Network (CNN), (ii) global movement tracking via acceleration data processed by the YoLov5 algorithm, and (iii) a two-layer Long Short-Term Memory (LSTM) model for final drowsiness assessment. The proposed approach achieves an average detection accuracy of 97.7%, demonstrating its effectiveness in preventing drowsy driving incidents.