IncorporationNet: a novel bimodal EEG-EOG vigilance estimation method via time-frequency-space feature fusion network

Area of Science:

• Neuroscience
• Biomedical Engineering
• Human-Computer Interaction

Background:

• Driver vigilance is crucial for road safety, requiring sustained attention and reaction capabilities.
• Electroencephalogram (EEG) and electrooculogram (EOG) signals are effective physiological measures for assessing vigilance.
• Existing methods may not fully leverage the complementary information from multimodal biosignals.

Purpose of the Study:

• To develop a bimodal feature fusion framework integrating EEG and EOG signals for enhanced vigilance estimation.
• To improve the accuracy and robustness of driver vigilance monitoring systems.
• To explore novel signal processing and machine learning techniques for fatigue detection.

Main Methods:

• A bimodal time-frequency-space feature fusion framework was proposed.
• Long Short-Term Memory (LSTM) networks combined with a Band-Spatial Attention Module (BSAM) were utilized.
• EEG sub-band dynamics and EOG temporal patterns were analyzed and fused via regression.

Main Results:

• The proposed framework achieved near state-of-the-art performance on the SEED-VIG dataset.
• The method demonstrated significant improvements in vigilance estimation accuracy, measured by RMSE and COR metrics.
• The fusion approach effectively reduced noise and enhanced the integration of multimodal features.

Conclusions:

• The bimodal vigilance monitoring approach offers a novel and effective methodology for driver attention assessment.
• This technique shows promising potential for real-time fatigue detection and driver safety applications.
• Integrating EEG and EOG signals via advanced feature fusion enhances the prediction of driver vigilance levels.