| Summary: | Deep learning has emerged as a powerful tool for developing Brain-Computer Interface (BCI) systems. However, the scarcity of subject-specific data results in a marginal performance increase for deep learning models trained entirely on the data from a specific individual. To overcome this, many transfer-based approaches have been proposed, using preexisting data from other subjects. But transfer learning faces its challenges: there are substantial inter-subject variabilities in electroencephalography (EEG) data. Therefore, adaptation is needed to fine-tune the model for the target subject. In this paper, we study 5 schemes for adapting a deep convolutional neural network (CNN) based EEG-BCI system for decoding hand motor imagery (MI). The proposed adaptation scheme improved the average accuracy of the base model by 1.93% (p<0.02). The adaptation resulted in an utmost of 35.18% increase in accuracy for a single subject, compared to a pre-trained base model.
|
|---|