Sliding Window Along With EEGNet-Based Prediction of EEG Motor Imagery

The need for repeated calibration and accounting intersubject variability is a major challenge the practical applications of brain–computer interface (BCI). problem becomes more challenging while decoding brain signals stroke patients due to altered neurodynamics caused by lesions. Recently, several deep learning architectures came into picture although they often failed produce superior accuracy compared traditional approaches mostly do not follow an end-to-end architecture as depend on custom features. However, few them have promising ability create generalizable features in fashion such popular EEGNet architecture. Although was applied patients’ motor imagery (MI) data, its performance good methods. In this study, we augmented EEGNet-based introducing postprocessing step called longest consecutive repetition (LCR) sliding window-based approach named it + LCR. proposed tested dataset ten hemiparetic MI yielding against only common spatial pattern (CSP) support vector machine (SVM) both within- cross-subject signals. We also observed comparable satisfactory LCR categories that are rarely found literature making candidate realize practically feasible BCI rehabilitation.