The use of optimal spatial filtering by the method of common spatial pattern for the classification of EEG signals according to the type of brain activity

Abstract

During EEG due to the volume conduction the signal from each individual source appears simultaneously in a number of channels registered from different leads. Therefore, the recorded EEG signal gives a blurred picture of human brain activity, which makes the task of classifying such signals rather complicated. One of the effective methods to obtain an informative signal from the EEG record is to use the optimal spatial filtering (some kind of deblurring), when the maximum content of the signal from the particular region of the brain (which corresponds to a certain type of brain activity) is achived in the output signal. An algorithm for classifying EEG signals using optimal spatial filtering by the method of common spatial pattern is proposed to identify two classes of brain activity -- imaginary left and right hand movements. To evaluate the quality of the algorithm, an EEG record, known as BCI Competition IV dataset 2b, was used. To determine the efficacy of the algorithm, the result of its operation was compared with the result of the algorithm without using optimal spatial filtering. The use of optimal spatial filtering improved the accuracy of classification from 0.74 to 0.79, which has shown its efficacy.