EEG Analysis of Sleepy Drivers: Identifying Key Brainwave Indicators of Fatigue

Abstract

Driver fatigue is widely recognized as a major cause of traffic accidents, resulting in millions of injuries and numerous fatalities each year. Existing driver fatigue detection systems primarily rely on facial cues and vehicle behavior. However, such approaches often fail to identify early signs of drowsiness. This study investigates the effectiveness of electroencephalogram (EEG) signals for early driver drowsiness detection using machine learning techniques. An open-access EEG dataset collected under controlled driving conditions using the NeuroSky MindWave headset was utilized. EEG features from multiple brainwave bands were analyzed to distinguish between alert and fatigued states. Power spectral density and band power analysis were employed for feature extraction, followed by feature importance evaluation and independent t-test to identify the most significant indicators. Three classification models—Random Forest (RF), Support Vector Machine (SVM), and Convolutional Neural Network (CNN)— were implemented and evaluated. Experimental results show that the RF model trained on a combined set of original and engineered band power features achieved the highest classification accuracy of 78.9%, outperforming CNN (74.0%) and SVM (72.8%). Future work will focus on validating the proposed approach using multi-channel EEG systems in real-world driving environments to enhance its practical applicability.