Sum of similarity-regularized squared correlations for enhancing SSVEP detection

Area of Science:

• Neuroscience
• Biomedical Engineering
• Signal Processing

Background:

• Brain-computer interfaces (BCIs) offer direct neural control of external devices.
• Steady-state visual evoked potential (SSVEP)-based BCIs are effective due to high information transfer rates (ITR) and minimal calibration.
• Existing SSVEP-BCI methods often overlook temporal dynamics and spatial coupling of EEG signals, and struggle with intrinsic noise.

Purpose of the Study:

• To develop a novel method for SSVEP detection that addresses limitations of existing approaches.
• To improve the accuracy and ITR of SSVEP-based BCIs.
• To reduce the calibration data requirements for effective SSVEP-BCI operation.

Main Methods:

• Proposed a novel method, Sum of Similarity-Regularized Squared Correlations (SSRSC), extending sum of squared correlations.
• Simultaneously computed squared correlations for calibration data and harmonic templates, mitigating variations via similarity regularization.
• Extended SSRSC using a ranking weighted ensemble strategy, termed weSSCOR.

Main Results:

• The proposed SSRSC/weSSCOR methods significantly improved SSVEP detection accuracy.
• Demonstrated enhanced information transfer rates (ITR) compared to existing methods.
• Achieved superior performance with reduced calibration data requirements on benchmark datasets.

Conclusions:

• SSRSC and weSSCOR are effective for SSVEP detection, outperforming current methods.
• These novel methods offer a promising approach for developing high ITR SSVEP-BCIs with reduced calibration needs.
• The techniques hold potential for practical applications requiring efficient and robust brain-computer interfaces.