Optimal and Suboptimal Reconstruction of Non Gaussian Realizations on the Output of Non Linear Converters

摘要

Two algorithms of sampling reconstruction procedures of non Gaussian processes are investigated. The optimal algorithm is analyzed on the basis of the conditional mean rule and cumulant functions. In this case the error reconstruction function depends on the given samples and on the probability density function of the given process. The non optimal algorithm is only based on the covariance function of the output process. We investigate the non Gaussian processes on the output of exponential and polynomial converters driven by the Gaussian Markov process. Comparison of both algorithms is given.