Fixed-point CCA algorithm applied to SSVEP based BCI system

Abstract

Canonical correlation analysis (CCA) has already been used to develop an on-line Steady State Visual Evoked Potentials (SSVEP) based Brain Computer Interface (BCI) system with high performance and stability. In this study, we proposed a fixed-point CCA algorithm which can be implemented in the embedded processors. It allows the implementation of low power-consumption portable BCI systems without PCs. It was mathematically proved that no overflow problem would occur during the entire process of this fixed-point algorithm. It was also shown that this algorithm could achieve a high calculation precision through the off-line SSVEP dataset. Finally, a number of on-line SSVEP based BCI experiments were conducted to demonstrate the speed of this algorithm. With a 240 MIPS processor, it merely cost 89ms for our algorithm to discriminate between 6 frequencies. The speed was fully compatible for the application of on-line SSVEP based BCI systems.