A toolbox for decoding BCI commands based on event-related potentials

Area of Science:

• Neuroscience
• Biomedical Engineering
• Computer Science

Background:

• Brain-computer interfaces (BCIs) commonly utilize event-related potentials (ERPs) for command decoding.
• Challenges in BCI development include selecting optimal EEG channels and features for classification.
• Existing methods often require significant programming expertise and complex feature extraction.

Purpose of the Study:

• To introduce a novel toolbox for automated ERP-based BCI decoding.
• To enable BCI control using a comprehensive set of EEG channels and automatically extracted features.
• To simplify the process of decoding brain activity for BCI applications.

Main Methods:

• Developed a toolbox for ERP-based decoding from electroencephalogram (EEG) data.
• Implemented automatic extraction of informative components from relevant channels.
• Utilized binary classification for handling sequences of stimuli, applicable to ERP-based spellers.
• Evaluated the toolbox on four diverse, openly available BCI datasets.

Main Results:

• The toolbox achieved performance comparable to state-of-the-art methods on multiple BCI datasets.
• Demonstrated successful application in P300-based spellers (matrix and RSVP), N2pc-based BCI, and error potential detection.
• Showcased the ability to handle complex stimulus sequences and multiple items via binary classification.

Conclusions:

• The developed toolbox reliably decodes ERPs for BCI applications with minimal user programming.
• Offers a user-friendly solution requiring only conventional preprocessing.
• Facilitates broader adoption and development of ERP-based BCIs.