球形阵列自适应网格分层定位算法

刚性球形传声器阵列可以在无空间模糊的条件下进行球谐域数据处理，基于球形阵列的可控波束形成器导向响应功率(SRP)算法定位精度高，但是计算量大，计算效率低。通过将球面致密的全局网格搜索替换为分层搜索策略可以有效减小SRP算法的计算量。提出MRE-SRP算法保持球形阵列SRP定位精度的同时降低计算量，首先通过球谐域MUSIC(SH-MUSIC)算法判断入射声源的数量减小搜索区域；其次将相对熵模型引入球谐域SRP(SH-SRP)定位算法中，提取网格分层前后的信息增益，设计自适应网格选择判据，实现分层多分辨率网格的精准再细分，从而降低计算量。实验验证了所提出算法的性能，结果显示在单双声源定位中，该算法可以实现较高的定位精度，精准选择分层网格，计算量减少75%以上。

Adaptive grid hierarchical localization algorithm based on spherical array

Rigid spherical microphone arrays can process spherical harmonic domain data without spatial ambiguity. Steerable response power (SRP) algorithm based on spherical arrays has high localization accuracy, but has a large amount of computation and low computational efficiency. The computational complexity of the SRP algorithm can be effectively reduced by replacing the spherical dense global grid search with a hierarchical search strategy. In this paper, the MRE-SRP (MUSIC and relative entropy joint localization SRP algorithm) algorithm is proposed to maintain the spherical array SRP positioning accuracy while reducing the amount of calculation. First, the spherical harmonic domain MUSIC (SH- MUSIC) algorithm is used to determine the number of incident sound sources and reduce the search area. Secondly, the relative entropy model is introduced into the spherical harmonic domain SRP (SH-SRP) positioning algorithm, the information gain before and after grid layering is extracted, and the adaptive grid selection criterion is designed to realize the accurate reproduction of layered multi-resolution grids subdivision, thereby reducing the amount of computation. Experiments verify the performance of the proposed algorithm, and the results show that in single and dual sound source localization, the algorithm can achieve high localization accuracy, accurately select the layered grid, and reduce the calculation amount by more than 75%.