Fatigue EEG signal detection based on Ensemble learning

In recent years, researchers have proposed a variety of machine learning and deep learning-based techniques to process EEG signals, such as support vector machines, deep learning algorithms, and graph neural networks (GNNs). Particularly, graph neural networks can effectively learn the spatial connectivity relationships between different channels in EEG signals, thereby enhancing the accuracy and robustness of fatigue driving detection. This paper introduces an integrated architecture that combines an information-enhanced network (IENet) with a graph convolutional network (GCN) to augment feature extraction capabilities and improve model classification performance. Through preprocessing EEG data and computing brain connectivity metrics, we trained the graph neural network and validated the method’s superior performance in fatigue driving detection. This research introduces novel methods and concepts aimed at advancing road traffic safety, particularly in mitigating the risks associated with fatigued driving.