Abstract: Double-layer-antenna beamforming can distinguish coherent sound sources at different parallel scan plane.However,its identification accuracy is unsatisfactory under low signal-to-noise ratio.To solve this problem,a modified algorithm is proposed:iterative regularization is used to update the regularization matrix and beamforming output.Thus,the side lobes are suppressed for the enhanced stability and the main lobe is sharpened to the real source point.Simulation and experiment results show that the proposed algorithm can locate the front and rear coherent sources with high accuracy.Therefore,modifying the original algorithm form the perspective of inverse problem is theoretically feasible.And the content of regularization matrix has tight relation with the spatial resolution of generalized inverse beamforming.These two factors can be integrated to improve the identification accuracy by iteration method.