Cross-Subject Driver Fatigue Detection Based on Attention-Guided Multi-Scale Convolutional Neural Networks Using EEG Signals

Fatigue driving is one of the main causes of traffic accidents worldwide. Research shows that around 20% of traffic accidents are closely related to driver fatigue. Therefore, establishing a reliable and accurate driver fatigue detection model is crucial to reducing traffic accidents and improving road safety. Physiological signals such as electroencephalograms (EEG) can more directly reflect the driver's physiological state, thus offering higher detection accuracy. However, most existing CNN-based methods focus only on single-scale EEG feature extraction, failing to capture fine-grained characteristics comprehensively. Moreover, many researchers concentrate only on single-channel EEG fatigue detection, which can lead to information loss or redundancy. To address these issues, this paper proposes a multi-scale convolutional neural network (MSCNN) guided by a channel attention mechanism (CAM). The MSCNN captures multi-dimensional features of EEG signals, while the CAM explores channel contributions, emphasizing more valuable information to complement the MSCNN's performance. Our method achieves satisfactory results on the SADT public dataset, with an average accuracy of 79.62% across 11 subjects, outperforming other fatigue detection methods. This demonstrates that our model is effective in EEG-based driver fatigue detection.