A Novel Algorithmic Structure of EEG Channel Attention Combined With Swin Transformer for Motor Patterns...

Area of Science:

• Neuroscience
• Biomedical Engineering
• Artificial Intelligence

Background:

• Electroencephalogram (EEG)-based brain-computer interfaces (BCI) are increasingly used in medical applications, particularly for stroke rehabilitation through motor imagery (MI).
• High-dimensional EEG data presents challenges for traditional deep learning models, often leading to overlooked channel correlations and classification errors.

Purpose of the Study:

• To develop a novel algorithmic structure for EEG channel-attention combined with Swin Transformer for enhanced motor pattern recognition in BCI rehabilitation.
• To address the limitations of traditional deep learning methods in handling high-dimensional EEG data and capturing intrinsic channel correlations.

Main Methods:

• Proposed a novel algorithmic structure integrating EEG channel-attention with the Swin Transformer architecture.
• Leveraged the self-attention mechanism within the transformer to capture temporal-spectral-spatial features in EEG data.
• Applied the method to motor pattern recognition for BCI rehabilitation.

Main Results:

• The proposed method achieved an average accuracy of 87.67%, outperforming existing state-of-the-art approaches.
• Demonstrated the ability to extract high-level, latent connections among temporal-spectral features in EEG data.
• Validated the effectiveness of channel-attention and Swin Transformer for high-performance BCI systems.

Conclusions:

• The combination of channel-attention and Swin Transformer offers a powerful approach for analyzing complex EEG data in BCI applications.
• This method significantly improves motor pattern recognition accuracy, showing great potential for advancing BCI-based neurorehabilitation.
• The findings suggest a promising direction for developing more effective and accurate brain-computer interface systems.