Range-Doppler Spectrum Estimation via Sparse variational Bayesian approach

Abstract

ABSTRACTIn this paper, the structural sparsity of the range profile and the range-Doppler (RD) image is utilised and an effective sparse variational Bayesian approach with modified automatic relevance determination (ARD) is proposed for the RD spectrum estimation. Specifically, a probabilistic model is derived where hierarchical sparse promoting prior is imposed over the scatter coefficients. Since the real and the imaginary parts of the same complex coefficients are strongly correlated, a symmetric construction hyper-parameter is designed. Furthermore, a new hyper-parameter learning rule is induced by optimising the marginal likelihood based on Majorization-Minimization (MM) principle to further improve the accuracy and the efficiency of the proposed algorithm. Finally, the simulation results demonstrate that the proposed algorithm can achieve superior performance and high efficiency in the scenarios of low signal-to-noise ratio (SNR) and limited pulse echoes.