Frequency Domain Filtering Method for SSVEP-EEG Preprocessing

Area of Science:

• Neuroscience
• Signal Processing
• Biomedical Engineering

Background:

• Steady-state visual evoked potential (SSVEP) signals are crucial for brain-computer interfaces but are often contaminated by noise.
• Traditional time-domain filtering methods for SSVEP signals have limitations in removing baseline drift and low-frequency interference.
• Accurate noise removal is essential for reliable SSVEP signal analysis and feature recognition.

Purpose of the Study:

• To propose and evaluate a novel frequency-domain filtering technique for SSVEP signals.
• To compare the effectiveness of the proposed method against traditional time-domain filtering.
• To enhance the performance of SSVEP feature recognition algorithms using the proposed filtering approach.

Main Methods:

• Empirical Mode Decomposition (EMD) was used to transform the time-domain SSVEP signal into a multi-dimensional signal.
• A 2-D Fourier transform was applied to convert the multi-dimensional signal to the frequency domain.
• A Gaussian high-pass filter was implemented in the frequency domain, followed by a 2-D inverse Fourier transform and signal reconstruction.

Main Results:

• The frequency-domain filtering method effectively suppressed baseline drift and low-frequency interference.
• Experimental results demonstrated significant improvements in feature recognition performance for Canonical Correlation Analysis (CCA), Filter Bank Canonical Correlation Analysis (FBCCA), and Task-Related Component Analysis (TRCA).
• The proposed method showed superior noise removal capabilities compared to time-domain filtering.

Conclusions:

• The proposed frequency-domain filtering method offers a significant advancement in SSVEP signal pre-processing.
• This technique enhances noise removal and improves the accuracy of SSVEP-based analysis.
• The findings suggest that this method can substantially promote the practical application of SSVEP systems.