A Novel Approach to Improve the Performance of the P300 Speller Paradigm

A brain computer interface (BCI) system provides a communication channel between a brain and a computer bypassing the need for muscular means. Electroencephalography based BCI systems that utilize the P300 speller paradigm are commonly used but due to the nature of the P300 speller paradigm, these systems are prone to erroneous classification. In this study, a novel approach addressing the variations in the characteristics of the P300 elicited by column and row flashes was proposed to overcome this problem. Markedly better results were achieved by both single trial analysis and averaging when the proposed approach was used. More than a 5% increase in single trial classification accuracy was observed, and the number of trials needed to obtain 100% classification accuracy was reduced by 44%. The results of this study warrant further research to implement the proposed approach in other methodologies.