EEG Based Driver Attention-Aware System for Safety

Abstract

The number of road accidents has drastically increased over the years, with a significant proportion attributed to driver drowsiness, fatigue, and distraction. Many of these accidents could be mitigated by issuing timely alerts to drivers when they exhibit signs of drowsiness or distraction. Elec-troencephalogram (EEG) signals offer a viable approach for monitoring and analyzing a driver's cognitive and physiological states. In this paper, we propose a method for estimating drivers' eye states utilizing EEG signals. We employed a dataset captured using a 16-channel EEG headset, emphasizing features extracted from the alpha, beta, gamma, and theta frequency bands through windowing techniques. Initial models based on logistic regression, decision trees, and random forests yielded suboptimal accuracy. However, by integrating features across multiple frequency bands and implementing a real-time analysis approach using a sliding window combined with a majority voting strategy, we enhanced the prediction accuracy to approximately 90%. This demonstrates the potential of EEG-based eye state prediction for enhancing driver drowsiness detection and intervention systems. © 2024 IEEE.