Robust single-trial ERP estimation based on spatiotemporal filtering.

Most spatiotemporal filtering methods for the problem of single-trial event-related potentials (ERP) estimation relies on the analysis of the second-order statistics (SOS) of electroencephalograph (EEG) data. Due to the noisy nature of EEG, these methods often suffer from the outliers in EEG. We combine a recently proposed spatiotemporal filtering method with the maximum correntropy criterion (MCC) for the single-trial estimation of the ERP amplitude. Study with real cognitive ERP data shows the robustness of the method with reduced estimation variance.