On Model Order Estimation in Distributed Passive Radar without Reference Signal

In this paper we consider a passive radar system with widely separated receivers where the propagation path between the target and some of the receiver elements may be blocked. We assume a system with no surveillance channel to the reference transmitted signal and consider the problem of joint detection of target and active receivers. We pose this as a composite detection problem with unknown signal parameters with unknown model order and derive penalty terms for Bayesian Information Criterion (BIC) and Exponentially Embedded Family (EEF) methods of model order selection. Simulation experiments are used to compare performance of the BIC, EEF, and Minimum Description length (MDL) based detectors. Results show higher performance of the BIC over EEF and MDL.