The feature extraction and estimation of a steady-state visual evoked potential by the Karhunen-Loeve expansion

Abstract

The study justifies using the Karhunen-Loeve expansion (KLE) for the feature extraction of a steady-state visual evoked potential (SSVEP), with the research results contributing to this subject area. The obtained parameters that sufficiently describe the study process allow constructing the information technology of visual analyzer diagnostics. It will help ophthalmologists reveal the characteristics of a disease and determine the right diagnosis. To obtain the research results, the Karhunen-Loeve expansion of the SSVEP was implemented at different stimulation frequencies, which helped determine the optimal number of the informative features: 6 Hz - 18, 8 Hz - 15, and 10 Hz - 12. The expansion results of two one-channel and one two-channel SSVEPs were compared to establish the fact of the channels' correlation effects on the number of the informative parameters. It has been proved that, taking into account the interference between the registration channels, it is possible to use fewer informative parameters for diagnostics. The obtained results will be used at the information technology of ophthalmologic diagnostics; the stage of their evaluation and the number of informative parameters are critically important because it all affects the accuracy and reliability of a diagnosis.