A Robust Asynchronous SSVEP Brain- Computer Interface Based on Cluster Analysis of Canonical Correlation Coefficients

Abstract

Brain computer interface (BCI) systems allow a natural interaction with machines, especially needed by people with severe motor disabilities or those whose limbs are occupied with other tasks. As the electrical brain activity (EEG) is measured on the user scalp in those systems, they are noninvasive. However, due to small amplitude of the relevant signal components, poor spatial resolution, diversity within users' anatomy and EEG responses, achieving high speed and accuracy at large number of interface commands is a challenge. It is postulated in this paper that the SSVEP BCI paradigm, combined with multichannel filtering can provide the interface robustness to user diversity and electrode placement. A cluster analysis of the canonical correlation coefficients (computed for multi-channel EEG signals evoked by alternate visual half-field LED stimulation) is used to achieve this goal. Experimental results combined with computer simulation are presented to objectively evaluate the method performance. © Springer International Publishing Switzerland 2014.