IMPROVING THE PERFORMANCE OF MOTOR IMAGERY EEG-BASED BCIS VIA AN ADAPTIVE EPOCH TRIMMING MECHANISM

Brain-Computer Interfaces (BCI) are rapidly evolving within both academia and industry necessitating urgent actions taken to further improve the signal processing module of such systems. Electroencephalography (EEG) is the leading choice for designing BCIs due to its affordability, convenience to use, and implementation simplicity. Motor Imagery (MI), which is merely the imagination of motory tasks without physically performing them, is one of the most common techniques within EEG-based BCIs. Despite the width of MI applications, researchers typically face two groups of subjects; Some subjects imagine repeating the requested movement during the response intervals (epoch), while some others might execute the mental imagination of the activity only once, and not necessarily consistently within equal intervals after the stimulus is presented. The paper focuses on this challenge and proposes an adaptive framework which finds the time interval of subject's concentration on the MI task through three different scenarios. The proposed framework trims the epochs in a way that the irrelevant information within each epoch is discarded as the remainder illustrates the maximum performance of the subject in each trial. The framework is benchmarked on datasets retrieved from BCI Competition III-IVa and the results exhibit significant improvement against its counterparts.