Control or non-control state: that is the question! An asynchronous visual P300-based BCI approach.

Area of Science:

• Neuroscience
• Biomedical Engineering
• Human-Computer Interaction

Background:

• Event-related potential (ERP)-based brain-computer interfaces (BCI) offer reliable synchronous communication.
• Synchronous BCIs are impractical for daily use due to potential selections when users are inattentive.
• Previous attention-aware asynchronous ERP-BCI research has yielded inconsistent results.

Purpose of the Study:

• To investigate novel approaches for detecting user engagement in asynchronous ERP-BCIs.
• To enhance the reliability and practicality of BCIs for everyday applications.
• To address the limitations of synchronous BCI systems.

Main Methods:

• Utilized electroencephalogram (EEG) signals, analyzing classifier output and frequency-domain features.
• Developed a hybrid approach combining classifier output and frequency-domain features for state detection.
• Evaluated state detection capabilities in various control scenarios using offline data from 21 healthy volunteers.

Main Results:

• The hybrid method significantly outperformed individual methods in user state detection.
• Achieved an average correct state detection accuracy exceeding 95% for an asynchronous P300-based BCI.
• Demonstrated the effectiveness of all introduced approaches in preventing involuntary selections.

Conclusions:

• All investigated methods effectively detect user state, preventing involuntary selections in asynchronous P300-BCIs.
• The hybrid approach combining classifier output and frequency-domain EEG features is the most effective strategy.
• These findings advance the development of more robust and user-friendly asynchronous BCIs.