Unified Online Adaptation Framework for Correlation Analysis-based Spatial Filtering Methods in SSVEP-based BCIs

Area of Science:

• Neuroscience
• Biomedical Engineering
• Signal Processing

Background:

• Online adaptation is crucial for user-friendly brain-computer interfaces (BCIs), but its application to steady-state visual evoked potential (SSVEP) recognition is limited.
• Existing methods like online multi-stimulus canonical correlation analysis (OMSCCA) for SSVEP spatial filter adaptation are effective but not broadly extensible.
• This limits the development of generalizable calibration-free BCI algorithms.

Purpose of the Study:

• To propose a unified online adaptation framework for correlation analysis (CA)-based spatial filtering methods in SSVEP recognition.
• To enable calibration-free, continuous updates of spatial filters for advanced filtering techniques.
• To enhance the development of user-friendly, zero-calibration SSVEP-based BCIs.

Main Methods:

• Extended the least-squares (LS) unified framework for online adaptation without pre-calibration, allowing continuous spatial filter updates.
• Introduced a cross-stimulus transfer method for adapting common impulse response and generating user-specific templates using limited online unlabeled data.
• Adapted three advanced spatial filtering methods to online paradigms using the proposed framework and validated through simulations.

Main Results:

• The unified framework effectively promotes the development of zero-calibration SSVEP-based BCIs.
• The proposed online adaptation methods improved recognition performance by over 12% compared to OMSCCA.
• Demonstrated the framework's generalizability for transforming calibration-based methods into adaptive solutions.

Conclusions:

• The proposed unified framework provides a generalizable approach for online adaptation in SSVEP recognition.
• This work facilitates the creation of more practical and user-friendly BCI systems by eliminating the need for pre-calibration.
• The enhanced performance and adaptability pave the way for wider adoption of SSVEP-based BCIs.