Classification of Hand Movements from EEG Signals using Machine Learning Techniques

Abstract

Brain-Computer Interface (BCI) is a communication system that enables individuals who cannot use their existing muscular system and nervous system because of various reasons to interact with the outside world. BCI enables the person to communicate with some electronic devices by interpreting brain activities. In this study, seven distinct hand movements were classified using the steady-state visually evoked potential (SSVEP) based BCI. EEG signals, which were recorded by researchers from the Autonomous University, were used. The classification was performed using algorithms of Naive Bayes, Extreme Learning Machine (ELM), and Support Vector Machines (SVM) by applying Autoregressive (AR), Hjorth, and Power Spectral Density (PSD) features which were extracted from EEG signals. As a result, classifier performances are 71-86% for AR features, 60-79% for Hjorth features, and 75-93% for PSD features by regarding feature extraction methods. On the other hand, classifier performances are 69-86% for Naive Bayes, 76-93% for ELM, and 60-82% for SVM by regarding classification algorithms. Among achieved accuracy performances, the best accuracy is 93% when the combination of PSD features and ELM algorithm is used.