An Approach to EEG based BCI for Motor Imagery using Explainable Transfer Learning

In this contribution we investigate possibilities of using neural network classification methods for EEG based BCI. We focus on image-based feature representation. The time-frequency based representation and its selected variants along with bi-spectrum were used. Set of more experiments including performance comparison with KNN and more feature mappings is provided. The explainability analysis for the neural networks approach is provided along with corresponding findings.