EEG-Based Frequency Domain Separation of Upward and Downward Movements of the Upper Limb

Abstract

ABSTRACTOne of the fundamental challenges encountered when implementing the Motor Imagery based Brain-Computer Interfacing (BCI) paradigm is accurately classifying the Electroencephalography (EEG) signals that originate due to the same joint movements. This emanates from the limited spatial proximity in the corresponding brain regions. Here, we explore the feasibility of distinguishing arm-reaching movements specific to the right hand using multiple frequency bands in EEG signals despite the limited spatial differentiation of induced potentials. To address this challenge, a channel averaging method was used combining six electrodes positioned in close proximity to the motor cortex, intending to isolate and enhance electromagnetic activity in the brain associated with arm movements. This study was further refined by focusing on two distinct frequency bands: mu (8-12Hz) and beta (12-30Hz), each associated with different cognitive and motor functions. The results of our study revealed promising outcomes across two classification methods. Utilizing the Support Vector Machine (SVM) classification method, our proposed approach achieved an average accuracy of 59.3% while the K-Nearest Neighbors (KNN) classification approach yielded an average accuracy of 61.63% in distinguishing between upward and downward movements of the right arm.