Leveraging Cross-Subject Transfer Learning and Signal Augmentation for Enhanced RGB Color Decoding from EEG Data

Area of Science:

• Neuroscience
• Machine Learning
• Biomedical Engineering

Background:

• Decoding neural patterns for RGB colors from electroencephalography (EEG) is crucial for visual brain-computer interfaces (BCIs).
• Inter-subject variability and limited data present significant challenges in EEG decoding.
• Existing methods struggle with data sparsity and individual differences.

Purpose of the Study:

• To investigate the efficacy of transfer learning and signal augmentation for improving RGB color decoding from EEG signals.
• To address challenges of inter-subject variability and limited data availability.
• To enhance the performance of deep learning models for EEG-based BCI applications.

Main Methods:

• Employed transfer learning for cross-subject information transfer, pre-training models on diverse subjects and fine-tuning on target subjects.
• Utilized signal augmentation techniques, including frequency slice recombination and Gaussian noise addition, to increase data diversity.
• Applied deep learning models like DeepConvNet (DCN) and Adaptive Temporal Convolutional Network (ATCNet) with attention mechanisms.

Main Results:

• Achieved a classification accuracy of 83.5% across all subjects on an EEG dataset.
• Demonstrated significant improvement in decoding performance compared to baseline methods.
• Showcased the effectiveness of the combined transfer learning and signal augmentation approach.

Conclusions:

• Transfer learning and signal augmentation effectively mitigate data sparsity and inter-subject variability in EEG decoding.
• The proposed methodology shows promising implications for advancing EEG-based classification and BCI applications.
• Improved accuracy and reduced variability highlight the potential for more robust and personalized BCIs.