Freeing P300-Based Brain-Computer Interfaces From Daily Recalibration by Extracting Daily Common ERPs

Area of Science:

• Neuroscience
• Biomedical Engineering
• Human-Computer Interaction

Background:

• Brain-computer interfaces (BCIs) often require daily recalibration due to day-to-day variations in brain signals.
• Event-related potentials (ERPs) are commonly used for BCI control, but their stability across days is a challenge.
• Frequent recalibration hinders the practical, daily use of BCIs.

Purpose of the Study:

• To investigate the impact of daily recalibration on P300-based BCI performance across multiple days.
• To develop and evaluate a novel method for mitigating daily recalibration needs in ERP-based BCIs.
• To enhance the long-term usability of BCIs for controlling everyday devices.

Main Methods:

• Implemented a P300-based BCI system for home appliance control over five days.
• Compared BCI performance using recalibration-based (RB) and recalibration-free (RF) decoders.
• Developed a sparse dictionary learning algorithm to extract daily common ERP patterns for RF decoders.

Main Results:

• The recalibration-free (RF) decoder showed reduced performance on subsequent days compared to the recalibration-based (RB) decoder.
• The proposed method, applied to the RF decoder, significantly improved performance on later days.
• The enhanced RF decoder's performance approached that of the RB decoder, reducing the gap caused by daily variations.

Conclusions:

• Daily recalibration poses a significant barrier to the widespread adoption of P300-based BCIs.
• The developed sparse dictionary learning method effectively extracts common ERP patterns, reducing the need for daily recalibration.
• This approach offers a promising solution for making BCIs more practical and accessible for daily life applications.