EEG motor/imagery signal classification comparative using machine learning algorithms

Electroencephalography (EEG) study allows the recording of brain activity associated with different mental tasks through electrodes placed on the scalp that amplifies the electricity changes at neurons activity. Because of the nature of EEG signals, their interpretation and classification require an expert. Recently, machine learning algorithms for EEG analysis have gained popularity and are applied in various activities such as brain-computer interfaces (BCI), diagnostic of brain disorders, etc. In this work, an EEG classification was performed with different machine learning algorithms. For this, Support Vector machines (SVM), K-nearest neighbor (KNN), Quadratic Discriminant Analysis (QDA), Linear Discriminant Analysis (LDA), Naive Bayes (NB) and Ensemble were implemented and the performance of different algorithms when distinguishing between two classes: one of movement and one of inactivity. The movement class was composed of Motor Imagery(MI) data and actual movement and inactivity class of a baseline. From the proposed techniques for EEG classification, the QDA and NB achieve the highest accuracy.