Multivariate weighted recurrent network for analyzing SSMVEP signals from EEG literate and illiterate

¹ School of Electrical and Information Engineering, Tianjin University - Tianjin 300072, China
² Department of Engineering Mathematics, University of Bristol - Bristol, BS1 5QD UK

Received: 5 July 2019
Accepted: 19 August 2019

Abstract

Recently, the Steady-State Motion Visual Evoked Potential (SSMVEP)-based Brain Computer Interface (BCI) has attracted a lot of attention. We design a SSMVEP-based BCI, in which a ring-shaped motion checkerboard pattern is used to realize SSMVEP stimulation. In particular, we firstly conduct SSMVEP experiments to obtain electroencephalogram (EEG) signals from 10 subjects, including 5 EEG literates and 5 EEG illiterates. By using the Canonical Correlation Analysis (CCA) and Support Vector Machine (SVM) method, we find that the classification accuracies of EEG illiterates are relatively lower than that of EEG literates. Thus, in order to investigate the differences in brain cognitive processes between the two groups of subjects, we construct a multivariate weighted recurrence network and analyze the weighted local efficiency and the clustering coefficient of the two groups. The results indicate that in SSMVEP experiment, there are significant differences between the two groups of subjects in these two network indicators. Our approach and analysis provide novel insights into the cognitive behavior of the brain and understanding of the “BCI Illiteracy” problem.