Robust sparse Bayesian learning for DOA estimation in impulsive noise environments

Abstract

Conventional direction of arrival (DOA) estimation methods are derived under Gaussian distributional assumptions on the noise and inevitably induce undesirable biases in impulsive noise environments. Therefore, in this paper, we propose a robust sparse Bayesian learning (SBL) method to correct potential outliers in observations and make full use of all observations for DOA estimation. Unlike the existing SBL methods, we model the measurements as a mixture of clean data and outliers to better learn the probability distribution of the observations. Motivated by the mixture model, we implement the variational Bayesian inference to alternately estimate the sparse signals and outlier noise. Then the impulsive noise components are subtracted from observations during the sparse signals recovery. Simulation results show that our method is superior to state-of-the-art techniques and can resolve coherent sources.