Classification of EEG Signals Using Alpha and Beta Frequency Power During Voluntary Hand Movement

Abstract

Pattern recognition using non-invasive techniques like electroencephalography (EEG) is valuable to infer and evaluate the neural interaction. In this study, EEG have been compared during the presence and absence of voluntary hand movement. Components of the alpha and beta frequency bands like the sensorimotor rhythm originated from the primary motor cortex and related brain areas reflect human movement. The power of 8–13 Hz alpha and 14–30 Hz beta frequency bands were used for the classification. To classify the data, k-NN algorithms (kNN), support vector machines (SVM), logistic regression (LR), decision tree classifiers (DT), linear discriminant analysis (LDA) and Gaussian naive bayes (NB) machine learning algorithms have been used. The best classification accuracy was achieved using decision tree algorithms which had an accuracy average f-score of 0.88 among four participants. In conclusion, decision tree classifiers ought to make alpha/beta frequency band based feature extraction for recognition of human movement.