An Online Adaptation Framework for Enhancing Calibration-Free SSVEP-Based BCI Performance

Area of Science:

• Neuroscience
• Biomedical Engineering
• Signal Processing

Background:

• Achieving plug-and-play steady-state visual evoked potential (SSVEP)-based brain-computer interfaces (BCIs) is challenging due to limitations in calibration-free decoding algorithms.
• Online adaptive canonical correlation analysis (OACCA) improves calibration-free performance via self-adaptation using online data, but its adaptation is limited to spatial filters, excluding other adaptive procedures like individual template estimation.
• This exclusion hinders fully exploitable model decoding and adaptation, necessitating more comprehensive online adaptation strategies.

Purpose of the Study:

• To propose and evaluate a novel online adaptation framework, online adaptive extended correlation analysis (OAECA), designed to enhance calibration-free SSVEP-based BCIs.
• To augment the online adaptation loop by incorporating individual template tuning and extended feature matching alongside spatial filter adaptation.
• To demonstrate the superiority of OAECA over existing methods like OACCA in terms of decoding accuracy and information transfer rate.

Main Methods:

• Developed the online adaptive extended correlation analysis (OAECA) framework, which includes recalling and cleaning online trials, tuning individual templates and spatial filters, and employing extended feature matching.
• Validated OAECA using two public SSVEP datasets for simulation experiments.
• Conducted both offline and online experiments to confirm the effectiveness and practical performance of the OAECA framework.

Main Results:

• OAECA significantly outperformed the state-of-the-art OACCA across almost all 105 subjects in simulation results.
• Both offline and online experiments confirmed the superior effectiveness of OAECA compared to OACCA.
• In online experiments, OAECA achieved a highest average information transfer rate (ITR) of 202.17 bits/min, surpassing OACCA's 177.02 bits/min.

Conclusions:

• The proposed OAECA framework offers a comprehensive online adaptation approach for SSVEP-based BCIs, significantly enhancing calibration-free decoding performance.
• OAECA's ability to tune both spatial filters and individual templates, coupled with extended feature matching, leads to substantial improvements over previous methods.
• This research advances SSVEP-based BCIs, moving them closer to practical, real-world plug-and-play applications.