Evaluating the Motor Imagery Classification Performance of a Double-Layered Feature Selection on Two Different-Sized Datasets

Numerous investigations have been conducted to enhance the motor imagery-based brain–computer interface (BCI) classification performance on various aspects. However, there are limited studies comparing their proposed feature selection framework both objective and subjective datasets. Therefore, this study aims provide a novel that combines spatial filters at frequency bands with double-layered evaluates it published self-acquired Electroencephalography (EEG) data preprocessed decomposed into multiple sub-bands, whose features then extracted, calculated, ranked based Fisher’s ratio minimum-redundancy-maximum-relevance (mRmR) algorithm. Informative filter banks chosen for optimal by linear discriminative analysis (LDA). The results of study, firstly, show method is comparable other conventional methods through accuracy F1-score. also found hand vs. feet more discriminable than left right (4–10% difference). Lastly, common pattern (FBCSP, without selection) algorithm be significantly lower (p = 0.0029, p 0.0015, 0.0008) compared when applied small-sized data.