Performance Improvement for Devanagari Script Input Based P300 Speller

Abstract

The performance of P300 speller-based BCI systems are evaluated based on its classification accuracy as well as the time required to spell the target character which is known as information transfer-rate (ITR). Though, previous works on English, Chinese and Devanagari script (DS) based P300 spellers have presented different approaches to improve the ITR. However, conventional machine learning approaches requires more number of trials for accurate detection of the target P300 component which increases the time to spell a character. The main aim of this work is to improve the information learning rate in DS based P300 speller by improving the classification accuracy in less number of trials. The deep convolution neural network was implemented for the detection of the target characters. The experimental results shown that the proposed work is able to provide higher accuracy for less number of trials which improved the ITR of existing DS based P300 speller.