Filter bank canonical correlation analysis for implementing a high-speed SSVEP-based brain-computer interface

Area of Science:

• Neuroscience
• Biomedical Engineering
• Signal Processing

Background:

• Canonical Correlation Analysis (CCA) is effective for steady-state visual evoked potential (SSVEP)-based brain-computer interfaces (BCIs).
• Existing CCA methods do not fully leverage harmonic SSVEP components for improved frequency detection.
• Enhancing SSVEP detection is crucial for advancing BCI performance and applications.

Purpose of the Study:

• To propose and evaluate a Filter Bank Canonical Correlation Analysis (FBCCA) method for SSVEP detection.
• To incorporate fundamental and harmonic SSVEP frequency components to boost CCA-based BCI performance.
• To compare different filter bank designs for optimizing FBCCA performance.

Main Methods:

• Developed FBCCA to integrate fundamental and harmonic SSVEP frequency components.
• Evaluated performance using a 40-target BCI speller with frequency coding (8-15.8 Hz).
• Compared three filter bank designs (M1, M2, M3) and standard CCA using offline data from 12 subjects and online tests with 10 subjects.

Main Results:

• FBCCA methods significantly outperformed standard CCA in SSVEP detection.
• The FBCCA method utilizing multiple harmonic frequency bands (M3) achieved the highest classification accuracy.
• An online BCI speller using optimal FBCCA demonstrated a high information transfer rate (ITR) of 151.18 ± 20.34 bits/min at ~33.3 characters/min.

Conclusions:

• The proposed FBCCA method significantly enhances SSVEP-based BCI performance by integrating harmonic components.
• FBCCA facilitates practical applications of BCIs, such as high-speed spelling.
• This approach represents a significant advancement in SSVEP detection for brain-computer interfaces.