Mixed near-field and far-field sources localization method using sparse Bayesian learning

To localize mixed near-field and far-field sources, this paper develops an algorithm based on sparse reconstruction theory. The proposed method takes full account of the correlation property between plane wave steering vectors and that of spherical waves. By creating over-complete dictionaries for the near-field and far-field areas separately and utilizing sparse Bayesian learning technique, the method reconstructs the space spectrum of the mixed sources successively. The separation and localization of mixed sources are completed at the same time, which can refrain the accumulative error caused by differencing approach of near-field and far-field sources. The proposed algorithm can deal with Gaussian signals and non-Gaussian signals without knowing the number of sources. Computer simulation results validate the effectiveness and the high precision of the proposed algorithm.