Preprocrssing of Event-Related Potential Signals via Kalman Filtering and Smoothing

In this paper an approach for solving the problem of event-related potential (ERP) identification, based on Kalman filtering and Kalman smoothing is presented. We assume that previous trials contain prior information relevant to the next trial and there are little dynamical changes from trial to trial. The results are presented for both simulated and real data. Simulated data were obtained by adding Gaussian functions with time-varying amplitudes and latencies, and real data were acquired during a common odd-ball type paradigm. The results show that this method has potential to denoise the ensemble averaged ERPs.