A method for estimating the parameters of the K-distribution using a nonlinear network based on fuzzy system and neural networks

This paper investigates a new technique for estimating the shape parameter of a K-distribution based on fuzzy neural network (FNN). In order to improve the estimation accuracy with inexpensive computational requirement, the FNN estimator is used to accurate the solutions of the nonlinear equations and the inverse functions (g_k(nucirc))of the Raghavanpsilas and ML/MOM (Maximum-Likelihood and Method Of Moments)methods respectively. A long this line, the estimated arithmetic and geometric means of data and the estimated function g_k (nucirc) of the two estimators are combined and modeled by the FNN shape parameter estimator where an off-line optimization of their weights via genetic algorithms (GA) is considered. The simulation results are carried out to demonstrate the validity of the approach as well as the successfulness of the FNN estimator for low mean square error (MSE) of parameter estimates when compared with existing Raghavanpsilas, HOFM (Higher Order and Fractional Moments), ML/MOM and [(z)log(z)] estimators. Additionally, the FNN method yields parameter estimates with lower computational complexity which allows rapid calculation in real time implementation.