Spatio-Temporal Filtering for Evoked Potentials Detection

Abstract

We propose the new application of the spatio-temporal filtering (STF) method, which is a detection of visual evoked potentials applied to brain-computer interfaces (BCI). STF aims in creating a new, enhanced channel basing on the current and the neighbouring samples from all the input channels. The new channel of the better quality facilitates quick detection of visual evoked potential in the EEG recording by reducing number of averaging operations. The BCI experiments include precise information on the times the specific events took place. This feature allowed us to design very accurately the learning step which is based on generalized eigendecomposition and aims in determining the spatio-temporal filter weights. STF based algorithm allows to achieve good results for enhancement and detection of visual evoked potentials applied for brain-computer interfaces. Advantageous classification accuracies obtained with the use of combined spatial and temporal approach suggest the method can contribute to improvement of the existing solutions and stimulate development of more accurate and faster EEG based interfaces between machines and humans.