Model order selection based on different information criteria for PDF estimation using maximum entropy method and application in cognitive radio systems

Abstract

In this paper, we propose two new information criteria to select the desired model order for probability density function (PDF) estimation using the maximum entropy method (MEM). These two proposed information criteria are based on Akaike information criterion (AIC) and Bayesian information criterion (BIC), respectively. The PDF estimation using MEM can be presented using integer and fractional moments. We use two proposed information criteria by considering trade-off between the goodness of fit of the model and the complexity of the model which result in obtaining the appropriate model order. In underlay cognitive radio (CR) systems, the primary user makes a powerful interference for the secondary user which changes the system noise PDF to a non-Gaussian one. The MEM can estimate this non-Gaussian PDF which in turn can be used in nonlinear detection schemes to suppress the degrading effect of the primary user. The simulation results show the high accuracy of the proposed model order selection criteria.