Abstract
Brain computer interface (BCI) represents a direct communication between the brain and any external device. P300 speller is one of the widely used BCI paradigms. BCI data are considered to be high in their dimensionality, which reduces the system performance. In this paper, we study the impact of adding feature selection techniques to the performance of P300 based BCI. Three types of feature selection techniques were applied and compared. These types are filter, wrapper, and hybrid methods. Fisher score, Determination Coefficient (r(2)), Regularized Fisher Linear Discriminant (RFLD), and Bayesian Linear Discriminant Analysis (BLDA) were used as evaluation functions. Differential Evolution (DE) optimization technique was used as searching technique. r(2) was preferred to be selected as feature selection method for P300 based BCI. This is due to the good reduction in dimension 64.8% and low computational cost 6.75 ms. The time required for training and testing the classifier was improved by 83.62%.