一种基于图神经与时间卷积的疲劳检测网络OA

Fatigue driving is a critical issue in the field of traffic safety,posing significant threats to lives and property due to traffic accidents caused by driver fatigue.Electroencephalogram(EEG)signals,which directly reflect the physiological state of drivers,have become an impor-tant data source for fatigue driving detection.However,existing EEG analysis methods often overlook the spatial correlations between brain re-gions and the temporal dynamics of EEG signals,limiting further improvements in detection performance.To address this,we propose a hy-brid model named EGT-Net,based on Graph Neural Networks(GNN)and Temporal Convolutional Networks(TCN),for fatigue driving de-tection.This model leverages GNN to capture the spatial topological relationships between brain regions in multi-channel EEG signals,while TCN is employed to extract temporal dependency features from EEG signals,enabling efficient classification of fatigue states.Experimental re-sults demonstrate that EGT-Net significantly outperforms traditional methods on the SEED-VIG dataset,achieving state-of-the-art accuracy and robustness.EGT-Net not only provides an efficient and accurate solution for fatigue driving detection but also advances the application of deep learning in EEG signal analysis,offering both theoretical significance and practical value.