Motor Imagery Events Classification by Utilizing Covariance and Correlation Matrices Generated from Electroencephalogram Signals

The brain-computer interface (BCI) enables a disadvantaged person to communicate with the outside world. Motor imagery (MI) is an important application of BCI that uses EEG (electroencephalogram) signals to capture brain activity while imagine of the movement. This paper uses correlation coefficient and covariance matrices to extract features from EEG signals obtained during motor imagery tasks and classify them using the Feed Forward Neural Network, Naive Bayes Classifier, and Linear Discriminant Analysis Classifier. The four categories of Dataset 2a of BCI Competition IV were examined. The feature obtained from the covariance matrix provided the highest accuracy for all types of classifiers used. The achievement of this study proves its potential use in BCI applications. They could restore paralyzed people's movements, improve the quality of life for people with disabilities, and open up new possibilities for human-computer interaction, such as the virtual world.