Classification of EEG-P300 Signals Extracted from Brain Activities in BCI Systems Using ν-SVM and BLDA Algorithms

Abstract

In this paper, a linear predictive coding (LPC) model is used to improve classification accuracy, convergent speed to maximum accuracy, and maximum bitrates in brain computer interface (BCI) system based on extracting EEG-P300 signals. First, EEG signal is filtered in order to eliminate high frequency noise. Then, the parameters of filtered EEG signal are extracted using LPC model. Finally, the samples are reconstructed by LPC coefficients and two classifiers, a) Bayesian Linear discriminant analysis (BLDA), and b) the υ-support vector machine (υ-SVM) are applied in order to classify. The proposed algorithm performance is compared with fisher linear discriminant analysis (FLDA). Results show that the efficiency of our algorithm in improving classification accuracy and convergent speed to maximum accuracy are much better. As example at the proposed algorithms, respectively BLDA with LPC model and υ-SVM with LPC model with8 electrode configuration for subject S1 the total classification accuracy is improved as 9.4% and 1.7%. And also, subject 7 at BLDA and υ-SVM with LPC model algorithms (LPC+BLDA and LPC+ υ-SVM) after block 11^th converged to maximum accuracy but Fisher Linear Discriminant Analysis (FLDA) algorithm did not converge to maximum accuracy (with the same configuration). So, it can be used as a promising tool in designing BCI systems.