Filter banks guided correlational convolutional neural network for SSVEPs based BCI classification.

Area of Science:

• Neuroscience
• Computer Science
• Signal Processing

Background:

• Steady-state visual evoked potential brain-computer interfaces (SSVEP-BCIs) utilize convolutional neural networks (CNNs) for effective signal processing.
• Traditional CNNs often rely on long time windows and frequency domain information, limiting performance in shorter durations.
• Existing methods may lack comprehensive task-related information beyond frequency domain data.

Purpose of the Study:

• To introduce a novel time-frequency domain generalized filter-bank convolutional neural network (FBCNN-G) for enhanced SSVEP-BCI classification.
• To address limitations of existing methods in short time windows and incorporate richer feature information.
• To improve the accuracy and efficiency of SSVEP-BCI systems.

Main Methods:

• Developed the FBCNN-G model integrating multiple EEG frequency information with sine-cosine signal priors.
• Employed filter banks divided into specific frequency bands as pre-filters for comprehensive feature extraction.
• Incorporated correlation analyses in both template and signal aspects for robust feature representation.

Main Results:

• The FBCNN-G model demonstrated superior character recognition accuracy and information transfer rates compared to other methods on the Benchmark dataset.
• Achieved a mean accuracy of 62.02%±5.12% in a short 0.2 s time window, highlighting its effectiveness.
• The model's performance was significantly improved across various time windows.

Conclusions:

• The FBCNN-G model offers a significant advancement in SSVEP-BCI classification performance, particularly in short time windows.
• This approach effectively integrates time-frequency information and prior signal knowledge for improved feature extraction.
• The FBCNN-G model is crucial for developing more efficient and accurate SSVEP-BCI character recognition systems.