Research on driving fatigue detection and arousal based on brain functional connectivity networks

Objective Driver fatigue poses a serious threat to road safety. This study presents a method for detecting driving fatigue and initiating wakefulness based on electroencephalogram (EEG) signals.Methods A total of 1,230 EEG samples were collected from 30 drivers during simulated driving. These samples were decomposed into θ, α, and β frequency bands using Discrete Wavelet Transform (DWT). A brain functional connectivity network was constructed based on the Phase-Lag Index (PLI) to extract features. CNN-LSTM, Transformer, and logistic regression models were trained to evaluate arousal effects under visual, olfactory, auditory single-modality, and multimodal conditions.Results Results showed that the β frequency-band dataset achieved the highest average accuracy (0.69), with the Transformer temporal classification model performing best (accuracy 0.76). All arousal protocols effectively alleviated fatigue (p < 0.05), with the multimodal visual,auditory,and olfactory approach yielding the strongest effect, reducing fatigue levels by an average of 2.633 points. The arousal effect was more pronounced at higher fatigue levels.Conclusions This study provides a theoretical foundation for fatigue monitoring and intervention in intelligent cockpits and autonomous driving systems.