Enhanced use practices in SSVEP-based BCIs using an analytical approach of canonical correlation analysis

Abstract

The search for a better understanding of the brain's anatomy and its functions on human actions has been a harsh yet very useful task, especially for brain-computer interface (BCI) engineering applications and medical diagnosis using signals from patients. Analyses involving electroencephalogram (EEG) signals processing have proven to be of great significance for developing this field of study. A widely used approach for this purpose is a BCI based on steady-state visual-evoked potentials (SSVEP), which, in general, are signals characterized by the brain's evoked response to visual stimuli modulated at a certain frequency. This work aims thus to propose a generalization of the correlation coefficient, which entails canonical correlation analysis (CCA), and verify its behavior under varying parameters to establish better use practices in BCI applications, comprising physiological, technical and operational factors. Also, it aims to analyze and compare signals from an SSVEP-based BCI to the results obtained from this generalization. The results show that new parameters can be introduced to better select the stimulus frequency and choose a specific BCI application; also, the analytical equation presents a good match with results obtained from real signals; at last, the final CCA equation can be written as a more general rule based on the sampling rate ratio, thus ensuring a higher flexibility and reliability for this technique. (C) 2019 Elsevier Ltd. All rights reserved.