Comparison of Extended and Unscented Kalman Filters applied to EEG signals

Abstract

For years the Extended Kalman Filter (EKF) has been the algorithm for non-linear systems due to its simplicity and suitability to real time implementations. Because of its shortfalls, however, the Unscented Kalman Filter (UKF) was introduced to be an algorithm which was more accurate. Since then researches have been conducted to investigate the suitability of both algorithms in different areas. This paper presents a comparison of the estimation quality for the two algorithms when applied to a real time system - Electroencephalography (EEG).