Classification of two motor imagery based on EEG signals in brain computer interface systems using LDA, SVM and GMM methods

One of the issues recently has been studied with the increasing number of accidents, problems with brain and spinal cord and also with medical engineering advances, is communication with the brain via computers. Devices that provide a bridge between the people and the outside environment using brain signals, called brain computer interface (BCI)-based devices. In this paper, at first some pre-processing such as filtering and removing artifacts of the data has been done. After pre-processing step in the process of feature extraction, we have extracted electroencephalography (EEG) samples obtained from motor imagery under a common spatial pattern (CSP). After feature extraction, classification turn arrives. In this section, we train model using machine learning techniques based on a certain number of trials and evaluate it. In this paper, we introduce machine learning algorithms, including linear discriminant analysis (LDA), support vector machine (SVM) and Gaussian mixture model (GMM) to classify two motor imageries and compare them. Three LDA, GMM and SVM for training dataset IVa of BCI competition III is used and the accuracy of each respectively 72.6%, 73.3% and 82.1% achieved that shows the superiority of SVM over other methods in terms of classification accuracy. We also evaluate three methods in term of time complexity and ultimately. We show that the SVM method has greater time complexity compared to both GMM and LDA methods.