Noise robust radar HRR target recognition based on Bayesian sparse learning

Abstract

A noise robust statistical model based on Bayesian sparse learning (BSL) is developed to characterize the complex-valued high range resolution (HRR) radar target signal, motivated by the problem of radar automatic target recognition (RATR).We assume a sparseness-promoting prior on the complex echoes from the scattering centers and a Markov dependency for the location of the dominant scattering center between consecutive HRR signals in the hierarchical Bayesian model. Considering the low signal-to-noise ratio (SNR) problem for a test sample, the statistical model trained under the high SNR can be updated to match the measured test sample and the corresponding recognition decision can be made based on the updated model. Efficient inference is performed via variational Bayesian (VB) for the proposed Bayesian sparse model. To validate the formulation, we present the experimental results on the measured HRR dataset for target recognition and signal reconstruction, and provide comparisons to some other statistical models for RATR.