MOTOR IMAGERY CLASSIFICATION FOR BCI USING STOCKWELL TRANSFORM, DEEP METRIC LEARNING, AND DCNN WITH MIXUP AUGMENTATION

Inter-individual EEG variability is a major issue limiting the performance of
Brain-Computer Interface (BCI) classifiers. However, most previous deep learning
(DL) models are still using the dataset of multiple subjects to train a single model due
to the limited augmentation techniques available for EEG signals, and the difficulty of
collecting large EEG datasets from each subject. Building a DL model that can be
trained on an extremely small EEG training set of a single subject presents an
interesting challenge that this work is trying to address. The deep metric learning
(DML) model is known for the ability to converge on a small dataset, however, this
kind of model has not been studied yet on BCI EEG signals.