Shape Parameter Estimation for K-Distribution Using Variational Bayesian Approach

Abstract

The sea clutter component in some of the radar and sonar signal models can be statistically characterized as following a K-distribution. This distribution has a shape parameter that is directly related to the number of scatterers. Hence, the estimation of this shape parameter is an important problem and is traditionally addressed using the maximum likelihood (ML), the method of moments (MoM) and their variants. A shortcoming of these methods is lesser accuracy in comparison to the theoretical CRB. In this paper, a variational Bayesian algorithm is proposed that provides both improved convergence and superior accuracy in comparison to the existing algorithms.