Modeling Strategies and Spatial Filters for Improving the Performance of P300-speller within and across Individuals

Abstract

In recent years, improving the performance of cross-individual brain-computer interfaces (BCI) has become a research hotspot. This paper proposes a within-individual model and two cross-individual models for P300 speller character recognition and uses canonical correlation analysis (CCA) spatial filter and task-related component analysis (TRCA) spatial filter to optimize the model. Those methods are compared in their performance to allow for an accurate classification of P300 speller. As a result, the best classification accuracy rate of the within-individual recognition model is 98.83%, and the best classification accuracy rate in cross-individual model is 85.09%.