基于非同步测量的高分辨率声源定位  

High-resolution sound source localization based on non-synchronous measurements

作　　者：韦娟[1] 冯鹏 宁方立[2] WEI Juan;FENG Peng;NING Fangli(School of Communication Engineering,Xidian University,Xi’an 710071,China;School of Mechanical Engineering,Northwestern Polytechnical University,Xi’an 710072,China)

机构地区：[1]西安电子科技大学通信工程学院,陕西西安710071 [2]西北工业大学机电学院,陕西西安710072

出　　处：《通信学报》2025年第2期123-135,共13页Journal on Communications

基　　金：国家自然科学基金资助项目(No.52475132);陕西省重点研发计划基金资助项目(No.2024GX-ZDCYL-01-16);航空科学基金资助项目(No.20200015053001);2023西安市重点产业链技术攻关基金资助项目(No.23ZDCYJSGG0006-2023)。

摘　　要：针对非同步测量声源定位方法在低信噪比条件下存在主瓣较宽、易受旁瓣干扰等问题,提出一种子空间逼近算法与截断函数波束成形联合的非同步测量声源定位算法。该算法首先对缺省互谱矩阵进行奇异值分解,通过截断阈值保留主要奇异向量构建低维子空间,继而将缺省互谱矩阵投影到子空间,寻找最优解来补全矩阵。补全后的互谱矩阵通过截断函数波束成形算法实现声源定位。仿真和实验结果表明,与基于核范数最小化及其衍生模型的算法相比,所提算法在低信噪比条件下能够有效减小主瓣宽度、抑制旁瓣数量,矩阵补全误差平均降低了17.6%、声源重构误差平均降低了27%,证明该算法具有良好的鲁棒性和抗噪性。Concerning the issues of broad main lobes and susceptibility to sidelobe interference in non-synchronous measurements of sound source localization methods under low signal to noise ratio conditions,a non-synchronous measurement sound source localization algorithm based on subspace approximation algorithm and truncated functional beamforming was proposed.Initially,the deficient cross-spectral matrix was performed using singular value decomposition,and the primary singular vectors were retained by a truncation threshold to construct a low-dimensional subspace.Subsequently,the deficient cross-spectral matrix was projected onto this subspace,and the optimal solution for matrix completion was sought.After completion,sound source localization was achieved by using a truncated function beamforming algorithm.Simulation and experimental results demonstrated that,compared with algorithms based on nuclear norm minimization and its derivative models,the proposed algorithm effectively reduces the width of the main lobes and the number of sidelobes under low signal to noise ratio conditions.The matrix completion error is reduced by an average of 17.6%,and the sound source reconstruction error is reduced by an average of 27%,proving the algorithm’s robustness and noise resistance.

关 键 词：非同步测量 声源定位 矩阵补全 子空间逼近 

分 类 号：TN912.3[电子电信—通信与信息系统]