Subject Independent Brain-Computer Interface Spelling System

The aim of this study is to introduce a novel P300 speller paradigm for facilitating communication of individuals with severe motor disabilities who preserve mental abilities but lack the physical abilities to speak. The study proposes integration of a three dimensional (3D) and column-only P300 speller paradigm with an artificial neural network (ANN) and involves generating population-level feature sets. Such an approach holds promise for removing the necessity to obtain subject-specific training data-sets and is especially practical for patients with motor impairments from whom collecting a training data-set prior to operating a BCI is not feasible.<o:p></o:p>

EEG data were collected from 110 healthy subjects while they were exposed to a 3D and column-only flashing paradigm. The offline classification model was built with ANN which yielded a 2-class classification accuracy of 99.40% ± 0.21. The performance of this offline model was tested in online sessions conducted with 30 new participants. An online character identification success rate of 95.41±% was achieved across all participants.<o:p></o:p>