End-to-end fatigue driving EEG signal detection model based on improved temporal-graph convolution network.

Area of Science:

• Neuroscience
• Artificial Intelligence
• Road Safety

Background:

• Fatigue driving is a major cause of traffic accidents.
• Physiological signal-based methods offer objective and accurate fatigue detection.
• Electroencephalography (EEG) signals are promising for real-time fatigue monitoring.

Purpose of the Study:

• To develop an end-to-end fatigue driving detection algorithm using EEG signals.
• To address limitations of traditional methods, such as training complexity and real-time performance.
• To improve the accuracy and efficiency of fatigue detection systems.

Main Methods:

• Proposed an algorithm named MATCN-GT, combining a multi-scale attentional temporal convolutional neural network (MATCN) block and a graph convolutional-Transformer (GT) block.
• The MATCN block extracts features directly from raw EEG signals.
• The GT block processes inter-electrode EEG signal features using graph convolutions and a Transformer module for long-range dependency capture.

Main Results:

• The MATCN-GT model achieved an accuracy of 93.67% on the SEED-VIG dataset, outperforming existing algorithms.
• The GT block improved accuracy by 3.25% compared to traditional graph convolutional networks.
• The MATCN block demonstrated higher accuracy across different subjects than existing feature extraction methods.

Conclusions:

• The proposed MATCN-GT algorithm offers a highly accurate and efficient solution for fatigue driving detection.
• The integration of temporal and graph convolutional networks with attention and Transformer modules enhances EEG signal analysis.
• This technology holds significant potential for improving road safety by mitigating fatigue-related accidents.