Brain Activity EEG-P300 Signal Categorization from LPC based Estimation of Signal using Fisher Linear Discriminant Analysis

Abstract

In this paper, a new approach is proposed to categorize the EEG-P300 signals of brain activity using fisher linear discriminant analysis (FLDA) based on smoothed signal achieved from linear predictive coding (LPC) filter. In order to overcome the overtraining of classifier caused by the noisy and non-stationary data, EEG signal is reconstructed using LPC model before applying to FLDA. Since, the brain signals are noisy, LPC model can highlight the signal features and reduce the irrelevant information. In this paper, classification accuracy, maximum bitrate and speed to achieve stability in maximum accuracy of classification are computed to compare performance of the proposed algorithm (LPC-FLDA) and FLDA. The implementation results show that the efficiency of the proposed method in terms of classification accuracy and convergence time to achieve stability in maximum accuracy is better than FLDA algorithm. As example at the proposed algorithm with 8 electrode configuration the S4 converges to the maximum accuracy after tenth Block while this happens for FLDA algorithm after thirteenth Block and the total classification accuracy for this person at proposed algorithm is improved as 0.8% than FLDA algorithm.