Classification of Motor Imagery Tasks Using Extreme Learning Machine and Filter Bank Common Spatial-Spectral Pattern

One of the most important limitations of the CSP method is that it only extracts spatial filters. In addition, CSP only uses a specific frequency band to derive spatial filters. Therefore, a combination of filter bank and spatial-spectral patterns (CSSP) was used to extract suitable features. Before feeding the features to the extreme learning machine (ELM) classifier, a common average reference (CAR) filter was applied on the selected channels to minimize artifacts related to inappropriate reference selection. The evaluation results showed that the proposed method achieved an average accuracy of ۸۵.۲۴% for all subjects in the dataset IVa.

Keywords:

Brain-computer interface , Motor imagery , Common spatial-spectral patterns , Common average reference , Extreme learning machine