Task-oriented EEG denoising generative adversarial network for enhancing SSVEP-BCI performance.

Area of Science:

• Neuroscience
• Signal Processing
• Machine Learning

Background:

• Electroencephalogram (EEG) signal quality is crucial for brain-computer interface (BCI) performance.
• Existing EEG denoising methods often neglect the impact on subsequent BCI tasks.
• Optimizing EEG denoising for specific BCI tasks is essential for practical applications.

Purpose of the Study:

• To develop and evaluate a novel EEG denoising method optimized for improving BCI task performance.
• To enhance the signal-to-noise ratio (SNR) of EEG signals by preserving task-related components and removing irrelevant noise.

Main Methods:

• Proposed an innovative task-oriented EEG denoising generative adversarial network (TOED-GAN).
• Utilized GAN's generator for signal decomposition and reconstruction, and discriminator for distinguishing clean from noisy signals.
• Employed canonical correlation analysis (CCA) for classification tasks in steady-state visual evoked potential (SSVEP) based BCI.

Main Results:

• TOED-GAN demonstrated superior performance in EEG noise removal and SSVEP-BCI accuracy improvement.
• Achieved accuracy improvements of 18.47% and 21.33% compared to baseline convolutional neural network methods.
• Validated on both public and self-collected datasets, confirming model robustness.

Conclusions:

• The proposed TOED-GAN is an effective EEG denoising method tailored for SSVEP tasks.
• This task-specific denoising approach significantly enhances the performance of SSVEP-based BCIs.
• TOED-GAN contributes to improving the practical applicability of BCIs in real-world scenarios.