Leveraging Transfer Superposition Theory for Stable-State Visual Evoked Potential Cross-Subject Frequency Recognition

Area of Science:

• Neuroscience
• Biomedical Engineering
• Signal Processing

Background:

• Steady-state visual evoked potential (SSVEP)-based brain-computer interfaces (BCIs) commonly require individual calibration, leading to time consumption, visual fatigue, and reduced usability.
• Existing spatial filtering methods aim to improve SSVEP detection by reducing spontaneous activity interference but necessitate personalized calibration.

Purpose of the Study:

• To develop a calibration-free, cross-subject frequency identification method for SSVEP decoding in BCIs.
• To enhance the usability and performance of SSVEP-based BCIs by eliminating the need for individual calibration.

Main Methods:

• A multi-channel signal decomposition model was constructed.
• A cross least squares iterative method was employed to create individual-specific transfer spatial filters and source subject transfer superposition templates.
• An ensemble cross-subject transfer learning method was proposed, integrating source-subject transfer, global transfer, and sine-cosine reference templates for SSVEP frequency recognition.

Main Results:

• The proposed method significantly outperformed existing methods (FBCCA, TTCCA, CSSFT) in offline tests on two public datasets.
• The method demonstrated direct applicability to online SSVEP recognition without requiring calibration.
• The algorithm exhibited robustness, a critical factor for practical BCI applications.

Conclusions:

• The developed cross-subject transfer learning method offers a viable calibration-free solution for SSVEP-based BCIs.
• This approach enhances BCI usability by removing the need for time-consuming and fatiguing calibration procedures.
• The robust and high-performing algorithm paves the way for more practical and accessible BCI systems.