Micro-Expressions for Safer Roads: Deep Learning-Based Drowsiness Detection in Truck Drivers

Abstract: Drowsy driving is a major cause of traffic accidents, especially among truck drivers who work long hours with limited sleep. Although contemporary driver monitoring systems have employed deep learning and computer vision, many depend on obvious signs like eye closure, yawning, or head nodding, which emerge only after considerable fatigue has developed. This research suggests a novel system that utilizes micro-expressions—slight, unintentional facial movements—as initial signs of drowsiness. Micro-expressions are more reliably identified than explicit signals, allowing for earlier intervention to avert accidents. We developed a deep learning model using a specialized dataset aimed at capturing micro-expressions and enhanced it through pre-processing and augmentation methods to boost generalization. The system underwent testing with truck drivers, who experience increased fatigue risks stemming from extended hours, repetitive routes, and substantial vehicle loads. Our findings show more than 95% accuracy, exceeding current methods and validating the system's ability to enhance road safety by detecting fatigue early.