Long Coherent Integration in Passive Radar Systems Using Super-Resolution Sparse Bayesian Learning

Abstract

Maximizing the coherent processing interval (CPI) is crucial when performing passive radar detection on weak signal reflections. In practice, however, the CPI is limited by the target movement. In this work, the extent of the range and Doppler migration effects occurring when using a long CPI to integrate the returns from an L-band digital aeronautical communication system (LDACS) based passive radar is studied. In particular, our simulations underline the extensive Doppler migration effect that arises even for non-accelerating targets. To this end, the Keystone transform and fractional Fourier transform techniques are combined with the standard passive radar processing to enable the compensation of both range and Doppler migration effects. This non-model-based approach is, however, shown to have limitations, in particular for low signal-to-noise ratios and/or multitarget scenarios. To address these shortcomings, a novel model-based framework that allows to perform joint target detection and parameter estimation is developed. For this, a super-resolution sparse Bayesian learning approach is employed. This technique uses a multitarget observation model, which accurately accounts for the underlying range and Doppler migration effects and provides super-resolution estimation capabilities. This is particularly advantageous in the LDACS case since the narrow bandwidth generally limits the separation of closely spaced targets. The simulation experiments demonstrate the effectiveness of the algorithm and the advantages it provides when compared to the standard migration compensation approach.