Cluster analysis of Canonical Correlation coefficients for the SSVEP based brain-computer interfaces

In this paper, a novel method for detecting steady-state visual evoked potentials (SSVEP) using multiple channel electroencephalogram (EEG) data is presented. Accurate asynchronous detection, high speed and high information transfer rate can be achieved after a short calibration session. Spatial filtering based on the Canonical Corelation Analysis method proposed in [1] is used for identifying optimal combinations of electrode signals that cancel strong interference signals in the EEG. Data from a test group consisting of 21 subjects are used to evaluate the new methods and to compare results to standard spectrum analysis approach. Conducted research, for different length signal segments and five visual frequencies, showed improvement of both classification accuracy and detection speed.