Motion Imagery-BCI Based on EEG and Eye Movement Data Fusion

Area of Science:

• Neuroscience
• Computer Science
• Biomedical Engineering

Background:

• Brain-computer interaction (BCI) often relies on Electroencephalogram (EEG) signals.
• Eye tracking has shown potential as a complementary BCI modality, particularly for visual perception and cognition tasks.
• Integrating multimodal data sources can enhance BCI performance.

Purpose of the Study:

• To propose and evaluate a method for fusing EEG and eye movement data in motor imagery (MI) tasks.
• To compare the classification accuracy of fused data against unimodal EEG or eye movement data.
• To assess the feasibility of using eye movement data to compensate for partial EEG data loss.

Main Methods:

• Developed a fusion method combining EEG and eye movement data features from motor imagery tasks.
• Tested the fusion method at both the feature and decision layers.
• Evaluated performance with reduced EEG data (50% of electrodes) combined with eye movement data.

Main Results:

• Fusion of EEG and eye movement data achieved higher MI classification accuracy than using either modality alone at the feature layer.
• Decision-layer fusion outperformed feature-layer fusion in classification accuracy.
• Combining reduced EEG data (50%) with eye movement data resulted in only a minor decrease in accuracy (1.07%) compared to using all EEG electrodes.

Conclusions:

• The proposed fusion method effectively enhances motor imagery based BCI performance.
• Eye movement data can serve as a viable compensatory measure for partial EEG data loss in BCI.
• This approach offers a valuable strategy for augmenting motor imagery BCI applications.