Multi-Class Motor Imagery Classification

The Motor Imagery (MI) classification task is a high dimensionmultivariate and complicated subject. In this respect, the originalsignals are analyzed and minimal unique features of the classes areextracted to facilitate accurate classification of the actions performed.The fusion of common spatial pattern, Fisher discrimination ratio, andfilter bank alongside the SVM and CNN-LSTM are incorporated toprovide accurate clustering. As a result and after extensive simulations,it is shown that the CSP+ FDR + CNN-LSTM setup more accuratelydifferentiates the classes.