Discrimination of Multiclass Motor Imagery-Based Brain-Computer Interface

Abstract

Motor imagery (MI) based on electroencephalography (EEG) is one of the methods that the brain-computer interface (BCI) system uses to identify the expected behavior through brain signals. In this study, we aimed to develop an algorithm that is capable of differentiating between 4 MI movements. To achieve this, the Data Set IIa A from BCI competition IV was used to test the algorithm. We used independent component analysis (ICA) to preprocess the signal and wavelet technique to decompose the obtained signal into the desired frequency bands. We then inserted these as common spatial pattern (CSP) input, maximizing the variance between 2 classes using the 1-versus-1 (OVO) technique. Afterward, the support vector machine (SVM) classifier is used to obtain the best possible separation between the 2 classes. The obtained result shows improvement in some significant subjects compared with a previous study of these techniques.