Analysis of mental and music tasks from EEG signals

In this study, EEG records of different mental, motor and music tasks from six healthy subjects were analyzed. A 64-channel Biosemi ActiveTwo System was used for recording during imagination of seven different tasks. Features were extracted with independent component analysis (ICA) and wavelet methods (CWT) and further they were classified by support vector machine (SVM) and artificial neural networks (ANN) classifiers. The purpose of this study is to decide on a feature extraction method for EEG-based BCI systems which is efficient and easy to implement and moreover to investigate the availability of music task for BCI applications. Using music tasks for BCI applications is firstly introduced in this study. Classification performance of the experimental data of mental and motor tasks was compared with scientifically proven BCI III data set. The accuracy of the system is tested with Kappa statistics. The results of the study showed that the NN classifier exhibited superior performance with the CWT method. Finally, it was proven that the music tasks with the high classification performance can be used for Brain Computer Interface (BCI) applications.