Gridless Bayesian Inference for DOA Estimation with Coprime Array

In this paper, we propose a gridless Bayesian inference based direction-of-arrival (DOA) estimation algorithm with coprime array to overcome the basis mismatch arising in conventional on-grid sparse signal reconstruction based algorithms. Under Bayesian framework, the distributions of DOAs are assumed as mixtures of von Mises models yielding the continuous parameterization of the direction information, while the signals power is governed by a Bernoulli-Laplace model turning the sources number estimation into binary sequence detection. The Bayesian model is formulated on the second-order virtual coarray signals and the variational Bayesian inference is performed to realized accurate DOA estimation with increased number of degrees of freedom (DOFs). Simulation results show the effectiveness of the proposed algorithm especially at low signal-to-noise ratio (SNR).