Driving Fatigue Detection Based on Hybrid Electroencephalography and Eye Tracking

Area of Science:

• Neuroscience
• Computer Science
• Human-Computer Interaction

Background:

• Electroencephalograph (EEG)-based methods show success in detecting driving fatigue.
• Unimodal data may be insufficient for optimal fatigue detection due to incomplete information.

Purpose of the Study:

• To propose a novel multimodal architecture combining EEG and eye tracking data to enhance driving fatigue detection.
• To improve fusion efficiency and feature representation for more accurate fatigue monitoring.

Main Methods:

• A hybrid architecture processing EEG and eye tracking data separately through encoders.
• Utilizing a cross-modal predictive alignment module and 1D attention modules for feature fusion.
• Employing a linear classifier for fatigue recognition based on fused features.

Main Results:

• Achieved 99.93% accuracy in intra-session evaluation.
• Demonstrated 88.67% accuracy in cross-session evaluation, outperforming EEG-only, eye tracking-only, DCCA, and DGCCA methods.
• Achieved 78.19% accuracy in cross-subject evaluation, surpassing single-modality and CCA-based multimodal methods.

Conclusions:

• The proposed multimodal method significantly enhances driving fatigue detection compared to unimodal approaches.
• The hybrid EEG and eye tracking architecture offers superior performance over existing multimodal techniques.
• This approach holds promise for improving driver safety systems through more robust fatigue monitoring.