Sound source localization of non-synchronous measurements beamforming with block Hermitian matrix completion

Abstract

In this work, the block Hermitian matrix completion (BHMC) method, a fast non-iteration algorithm to complete the cross-spectral matrix, is proposed to realize the sound source localization of the non-synchronous measurements beamforming. Compared with the conventional beamforming (CBF) method of single measurement and the augmented Lagrange multiplier (ALM) method of non-synchronous measurements, the proposed method can obtain source maps with narrow mainlobes and no sidelobes. The simulation results show the matrix completion error and the sound source reconstruction error of BHMC are less than those of ALM. In terms of computational efficiency, the proposed method is more efficient than ALM. The computation time of BHMC increases slowly with the number of the non-synchronous measurements while that of ALM increases sharply. Finally, experiments were conducted to verify the feasibility of BHMC. The experimental results show that the proposed method is an accurate and fast non-synchronous measurements beamforming algorithm, which paves a way to the real-time sound source localization in industrial occasions.