Multipole transfer matrix model-based sparse Bayesian learning approach for sound source identification

Accurate source identification is key to solving complex flow-induced noise problems, and sources often have non-uniform radiation patterns. At the same time, traditional methods have many application limitations, including the need for prior knowledge of the sound field, as well as inability to apply to coherent sources or coexisting multipoles. A multipole transfer matrix model-based sparse Bayesian learning approach for sound source identification (MT-SBL) is proposed. In this algorithm, the transfer relationship for the multipole radiation patterns of sound sources is established. Then, the signal sparse reconstruction is transformed into an iterative updating problem of the feature parameters under a sparse Bayesian learning (SBL) framework, which leads to the accurate identification of multipole sources. Numerical simulations and experiments on electric speakers in an anechoic environment validate the proposed algorithm. Compared with the existing methods, the algorithm is more advantageous in terms of accuracy, resolution, and computational efficiency. Further experiments were conducted on the identification of supersonic jet noise sources, verifying the effectiveness of the method for identifying strongly directional sound sources.