Leveraging low-frequency components for enhanced high-frequency steady-state visual evoked potential based brain...

Area of Science:

• Neuroscience
• Biomedical Engineering
• Computer Science

Background:

• High-frequency steady-state visual evoked potential-based brain-computer interface (SSVEP-BCI) systems offer user comfort but have lower performance than low-frequency systems.
• This performance gap limits the practical application of high-frequency SSVEP-BCI.

Purpose of the Study:

• To enhance the performance of high-frequency SSVEP-BCI systems.
• To address the limitations of reduced performance and extensive calibration in high-frequency SSVEP-BCI.

Main Methods:

• A transfer learning approach is proposed, utilizing low-frequency SSVEP data to improve high-frequency SSVEP performance.
• A filtering mechanism extracts informative components from low-frequency signals.
• The least squares algorithm generates synthetic high-frequency data.

Main Results:

• Significant performance improvements were observed on two public datasets using TDCA, eTRCA, and TRCA-based algorithms.
• Accuracy increased by 9.03% and 14.49% for eTRCA and TDCA on Dataset 1, and 13.91% and 14.53% on Dataset 2, all within 1.5 seconds.
• The method requires only two calibration trials, effectively addressing the issue of single calibration data.

Conclusions:

• The proposed transfer learning method significantly improves high-frequency SSVEP-BCI performance.
• Fast calibration and enhanced accuracy are achievable for real-world high-frequency BCI applications.
• This approach makes high-frequency SSVEP-BCI systems more practical and user-friendly.