A Dictionary Learning-Based Off-Grid DOA Estimation Method Using Khatri-Rao Product

Abstract

Grid mismatch is the main drawback in grid-based sparse representation. For DOA estimation, off-grid problem degrades the accuracy of angle estimation. In order to solve this problem, a dictionary learning-based off-grid DOA estimation method is proposed. Firstly, we calculate the sampling covariance matrix, then based on covariance matrix model, we formulate the DOA estimation as a sparse representation problem with Khatri-Rao product dictionary. In the proposed method, two stage iteration strategy is utilized to address the off-grid problem. In the first stage, the coarse estimation is attained by the grid-based sparse DOA estimation; in the second stage, the dictionary perturbation parameter is learned based on gradient descent method for improving the accuracy of DOA estimation. Simulation results verify the effectiveness of the proposed method.