Visual P300 stimulator with non uniform background and non symmetric stimulation markers

Abstract

This paper presents the development of a visual stimulation screen prototype for P300 BCI based on the native OpenViBE Donchin speller where an RGB snapshot of the navigation scenario for an electric wheelchair was embedded on the background instead of the usual solid color. The letters which are the stimulation markers were appropriately replaced by a eyefish character and placed on landmarks reachable by the wheelchair. Eight subjects were recorded using a copy-mode oddball paradigm whose global target coherent average plots show P300 elicitation even with no standard background or target arrangement. Linear Support Vector Machines (linear-SVM) were trained for the automatic detection of P300 and evaluated with cross-validation to report Area Under Receiver Operating Characteristic Curve (AUROC) as a classification performance measure. Despite of the strong OpenViBE Donchin speller customization constraints, an acceptable P300 detection rate was reached, in the best case 0.91±0.06 AUROC. Several improvements of the prototype are discussed for future work.