Key points are not available for this paper at this time.
Most of the BCI systems need EEG data with several channels to reach good accuracy. However, exceedingly increasing the channel need will increase the amount of calculation, and in some cases, decrease the accuracy and will also make the implementation of a BCI system difficult. Therefore, identifying the most effective channels in BCI systems is crucial because it will decrease the complexity and increase system accuracy. The Generalized Sequential forward selection (GSFS) method is used in this paper to choose the channel in a motor imagery BCI system for classification of right and left hand. Firstly, data is filtered to be in the frequency range of 4-30 Hz because the results of previous research revealed that the highest effect of motor imagery is exerted inside this frequency range. The Common Spatial Pattern (CSP) features and frequency domain features are simultaneously used in order to improve the system performance. Moreover, a PCVM classifier is used to enhance the classification performance. Employing the GSFS method and also simultaneously extracting the CSP and frequency domain features have increased the system output accuracy. The computation cost of this method is low compared to that of the genetic algorithm method for channel selection. The classification precision in the method used in this research is higher with respect to that of the SVM-RFE method which shows the advantage of this method over other methods for channel selection in an MI-BCI system.
Radman et al. (Tue,) studied this question.