Multivariate Synchronization Index Based on Independent Component Analysis for SSVEP-Based BCI

Abstract

A template-matching approach combined with multivariate synchronization index (MSI) and independent component analysis (ICA) based spatial filtering for steady-state visual evoked potentials (SSVEPs) frequency recognition is proposed in this paper to enhance the performance of SSVEP-based brain-computer interface (BCI). As a type of electroencephalogram (EEG) signals, SSVEPs generated from underlying brain sources is different from other activities and artifacts, this spatial filter has great potential to enhance the signal-to-noise ratio (SNR) of SSVEPs. This study adapted the MSI-ICA based spatial filters to process test data and the averaged training data, and then used the correlation coefficients between them as features for SSVEP classification. Some conventional methods such as canonical correlation analysis (CCA), filter bank-CCA (FBCCA), and ICA based frequency recognition were adapted to do the contrasting experiment, using a 40-class SSVEP benchmark datasets recorded from 35 subjects. The experimental results demonstrate that the MSI-ICA based method outperforms other methods in terms of the classification accuracy and information transfer rate (ITR).