BIAS PARAMETER ESTIMATION FOR GENERALIZED GAUSSIAN FAMILY OF DISTRIBUTIONS

D.A. Kurkin, A.A. Roenko, V.V. Lukin

A family of Generalized Gaussian Distributions (GGD) is considered as a model for noise approximation. A method based on percentile coefficient of kurtosis and median absolute deviation estimation is proposed and its effectiveness for GGD is investigated. Median, myriad and meridian estimators are considered and experiments to investigate their effectiveness in bias estimation of GGD are carried out. Analysis of the experimental results allows determining applications of the considered estimators.

Keywords: generalized Gaussian distribution, percentile coefficient of kurtosis, absolute median deviation, median, myriad, meridian.