Motor Imagery Classification for Asynchronous EEG-Based Brain-Computer Interfaces.

Area of Science:

• Neuroscience
• Biomedical Engineering
• Machine Learning

Background:

• Motor imagery (MI) based brain-computer interfaces (BCIs) offer direct control via imagined movements.
• Asynchronous BCIs, unlike triggered systems, aim for seamless MI detection without explicit cues, posing significant implementation challenges.
• Current BCIs struggle with accurately distinguishing resting states from MI trials in an asynchronous, untriggered environment.

Purpose of the Study:

• To develop and validate a novel approach for motor imagery (MI) detection in asynchronous brain-computer interfaces (BCIs).
• To enhance the accuracy and efficiency of MI-based BCIs by introducing a robust classification strategy.
• To overcome the challenges of distinguishing between resting states and MI trials without external triggers.

Main Methods:

• Proposed a sliding window prescreening and classification (SWPC) approach for asynchronous MI-BCIs.
• Implemented a two-module system: a prescreening module for MI trial detection and a classification module for task identification.
• Employed a training strategy combining supervised and self-supervised learning to refine feature extractors.

Main Results:

• The SWPC approach demonstrated consistent effectiveness across four different EEG datasets.
• Achieved the highest average classification accuracy in both within-subject and cross-subject asynchronous MI classification tasks.
• Outperformed the leading state-of-the-art baseline methods by approximately 2% in classification accuracy.

Conclusions:

• The SWPC method represents a significant advancement for MI-based asynchronous BCIs.
• This approach effectively addresses the challenge of triggerless MI detection and classification.
• SWPC offers improved performance and robustness, paving the way for more intuitive BCI applications.