Deep Learning Algorithms for EEG-Based Driver Drowsiness Detection: A Comparative Investigation

Due to the significant variability in electroencephalogram (EEG) signals across individuals and recording sessions, developing a calibration-free system for detecting driver drowsiness presents a considerable challenge. While numerous deep learning techniques have been employed to analyze EEG signals for this purpose, most existing studies rely on deep learning models with limited success in accurately detecting drowsiness. In this research, we explored the application of five deep learning algorithms such as Artificial Neural Networks (ANN), Convolutional Neural Networks (CNN), Recurrent Neural Networks (RNN), Gated Recurrent Units (GRU), and Long Short-Term Memory (LSTM), to improve the detection of driver drowsiness. Our applied models achieve detection accuracies of 73.46% for ANN, 80.09% for CNN, 65.74% for RNN, 83.95% for GRU, and 77.62% for LSTM across 11 subjects, outperforming conventional baseline approaches and state-of-the-art deep learning methods. Notably, the GRU model demonstrated superior accuracy in recognizing drowsiness from EEG signals compared to other applied models.