Enhancing the performance of SSVEP-based BCIs by combining task-related component analysis and deep neural network.

Area of Science:

• Neuroscience
• Computer Science
• Biomedical Engineering

Background:

• Steady-State Visually Evoked Potential (SSVEP) signals are crucial for brain-computer interfaces (BCIs).
• Traditional machine learning and deep learning networks offer distinct advantages for SSVEP decoding.
• An efficient integration strategy for these methods in SSVEP BCIs is currently lacking.

Purpose of the Study:

• To propose and validate an efficient framework for combining traditional and deep learning methods for SSVEP signal recognition.
• To enhance the performance of SSVEP-based BCIs by leveraging the complementary strengths of different algorithms.

Main Methods:

• A novel classification framework, eTRCA+sbCNN, is proposed, integrating ensemble task-related component analysis (eTRCA) and a sub-band convolutional neural network (sbCNN).
• Both eTRCA and sbCNN models are trained independently.
• Classification score vectors from both models are summed, and the SSVEP frequency with the highest summed score is selected.

Main Results:

• The eTRCA+sbCNN framework significantly improves classification performance compared to individual eTRCA and sbCNN models.
• Validation on two SSVEP BCI datasets demonstrates superior performance against state-of-the-art methods.
• The proposed method effectively utilizes the complementarity of feature signals from both approaches.

Conclusions:

• The eTRCA+sbCNN framework offers an effective strategy for integrating traditional and deep learning methods in SSVEP BCIs.
• This approach significantly enhances SSVEP classification accuracy, promoting the practical application of SSVEP-BCI systems.