Motor Imagery based EEG Signal Classification using Electrode Optimization Methodology

Abstract

This paper focuses on motor imagery Electroencephalography (EEG) signal classification based on electrode optimization. Motor imagery (MI) signals captured through EEG is the popular non-surgical way of acquiring brain signals. MI-based Brain-Computer Interface (BCI) assists motor impaired persons to link to the outside world by undertaking a series of motor functions. The BCI system commonly includes filtering of raw brain signal, extraction of significant features and classification. The electrode optimization technique is used by selecting limited electrodes attached to the brain parts related to motor functioning. Preprocessing using band pass filtering is done between 7-30 Hz as mu (µ) and beta (?) patterns responsible for imagery movements lie within this frequency range to remove artifacts. To obtain excellent classification performance, preprocessing plays a vital role. Band pass filtering applied to selected channels gives the classification accuracy of 87.27% as compared to 66.39% obtained without filtering. Hence there is a 21% increase in accuracy with filtering using Linear Discriminant Analysis (LDA) as a classifier. The proposed system is validated using test dataset IVb of BCI competition III. The findings prove that the proposed system enhances accuracy for the selected channel of interest, thus reducing computational complexity.