A New Object-oriented SSVEP-based BCI Paradigm Using Continuous Action Scene

Abstract

In recent years, brain-computer interfaces (BCI) based on steady-state visual evoked potentials (SSVEP) have been utilized widely in applied fields, such as the speller, wheelchair control, prosthesis control, industrial robot and so on. Various paradigms were designed to improve BCI performance. However, the relationship between control objects and SSVEP paradigm has been neglected. In this study, we proposed a new object-oriented SSVEP-BCI paradigm. This paradigm used continuous action scene of controlled object to replace traditional stimulus to stimulate the generation of SSVEP, which improve SSVEP recognition accuracy and realize the more user-friendly BCI system of “what you see is what you get”. The SSVEP-BCI system for controlling a 2DOFs prosthesis hand was customized as an example and four healthy subjects were recruited for our experiments. Firstly, Electroencephalogram (EEG) data was analyzed in frequency domain and time-frequency domain. Results show that significantly strong SSVEP was elicited. Then canonical correlation analysis (CCA) algorithm was used for SSVEP recognition. Experimental results show that the mean recognition accuracy and ITR (mean ± standard deviation) in short time window (1-2s) reached $87.66\pm 2.09\%$ and $28.51\pm 5.63$ bits/min respectively, which was comparable or even better than the performances of conventional four-choice SSVEP-based BCIs. Finally, the paradigm is universal and can be applied to various SSVEP-BCI fields. Taken together, these results suggest that the proposed paradigm is a promising option in BCI applications.