压缩波束形成声源识别的改进研究

凭借空间分辨率高、旁瓣衰减能力强等优势，压缩波束形成声源识别算法备受关注。传统方法直接最小化声源分布向量的 范数，重构声源分布与真实声源分布之间存在一定偏差，声源无法被直接准确量化。为改善该问题，本文给出迭代重加权 范数最小化方法，其迭代求解声源分布，且每次迭代中对声源分布向量进行加权。仿真及试验结果均证明：所给方法能有效降低传统方法的重构偏差，能直接用主瓣峰值准确量化声源强度，且空间分辨率更高、旁瓣衰减能力更强。
Abstract：Compressive beamforming acoustic source identification algorithm receives much attention due to its high spatial resolution and strong side-lobe attenuation ability.The conventional method directly minimizes the l-norm of the acoustic source distribution vector to reconstruct the acoustic source distribution.There is a certain deviation between the reconstructed acoustic source distribution and the actual one to make acoustic sources not be directly and accurately quantified.To solve this problem, an iterative reweighted l-norm minimization method was proposed.With it, the acoustic source distribution was solved iteratively.The acoustic source distribution vector was weighted in per iteration.Simulation and test results demonstrated that the proposed method can effectively reduce the reconstruction deviation of the conventional method, and directly and correctly quantify the acoustic source intensity with main-lobe peaks; it has higher spatial resolution and stronger side-lobe attenuation ability.