Source Localization of Joint Subspace and Sparse Bayesian Learning in the Near Field

Source localization in the near field based on L1-regularization least square minimization involve some problems such as requiring to select the regularization parameter, large computation complexity and lower estimate accuracy. In order to solve these problems, a near field source localization algorithm is proposed by combining subspace and sparse Bayesian learning in this paper. Through separating the two parameters of bearings and ranges, the two-dimensional parameter estimation problem is first transformed into one-dimensional parameter estimation one. The sparse Bayesian learning theory is used to calculate the bearings efficiently then. After that, the problem of ranges estimation can be solved by subspace method. Compared with the L1-regularization least square minimization method, the proposed method does not needs to choose regularization parameter and shows lower estimate errors. In addition, simulation results are given to illustrate the effectiveness and efficiency of the proposed method.