A Robust Classification Approach for Character Detection Using P300-Based Brain-Computer Interface

Abstract

Researchers have been contributing to the brain-computer interface (BCI), which acts as a direct connection between the human brain and the computer that uses the P300 speller paradigm to decode the response of the brain by stimulating a subject, involving no muscular movements. This research uses BCI Competition III Dataset II, which uses a 6x6 character matrix paradigm for data collection purposes for two healthy subjects. The ensembles of support vector machine (SVM) method of classification has been proposed to surpass the problem of false detection, which is preceded by empirical mode decomposition (EMD) as the preprocessing technique and the use of a stacked autoencoder for feature extraction and covariate shift adaptation by normalized principal components as the feature selection method for better accuracy of the detected character. The experiment yields a better result than many existing methods; it produces an average accuracy of 98.75%.