An improved multiple signal classification method for a circular acoustic vector sensor array in the presence of model errors

Aiming at the problem that the direction-of-arrival (DOA) estimation performance of multiple signal classification (MUSIC) algorithm of acoustic vector sensor (AVS) uniform circular array is deteriorated by the non-normal covariance matrix, an improved MUSIC algorithm based on singular value decomposition (SVD) of combined processing method of pressure and particle velocity (PV-CPM) is proposed. The influence of array response error and noise model error on the normality of covariance matrix and the estimation performance is analyzed. The covariance matrix of AVS array is no longer a normal matrix under the condition of model error. By using SVD of the covariance matrix for PV-CPM, the singular vectors are used to span the noise subspace based on the orthogonality of the singular vectors of the non-normal matrix. Numerical simulation results show that the proposed method improves the DOA estimation accuracy and multi-target resolution ability with a lower and flatter spatial background. The lake experiment further verifies the effectiveness of the proposed method.