Parameter estimation in k-distributed clutter with noise using nonlinear networks

Abstract EN

compound K-distribution in presence of thermal noise by using the artificial neural networks (ANN) and the fuzzy neural networks (FNN) multi inputs / single outputs (MISO) approaches.

In order to accurate more the estimation, the FNN multi inputs / multi outputs (MIMO) is also considered in this work to estimate simultaneously the shape and the scale parameters of the compound K-distribution.

Using the back propagation (BP) algorithm and the genetic learning algorithm (GA), the proposed estimation procedures are trained based on the moments and the ratio of arithmetic and geometric means of the samples.

However, the estimation accuracy obtained by these techniques are validated and compared for various values of the shape parameter with different sample sizes.