分裂增广拉格朗日收缩反卷积声源识别算法  

Split augmented lagrange contraction deconvolution algorithm for sound source identification

作　　者：樊小鹏 张鑫 褚志刚[2,3] 李丽[1] FAN Xiaopeng;ZHANG Xin;CHU Zhigang;LI Li(Electric Power Research Institute,Guangdong Power Grid Co.,Ltd.,Guangzhou 510082,China;State Key Lab of Mechanical Transmissions,Chongqing University,Chongqing 400044,China;College of Automotive Engineering,Chongqing University,Chongqing 400044,China)

机构地区：[1]广东电网有限责任公司电力科学研究院,广州510082 [2]重庆大学机械传动国家重点实验室,重庆400044 [3]重庆大学汽车工程学院,重庆400044

出　　处：《振动与冲击》2020年第23期141-147,共7页Journal of Vibration and Shock

基　　金：中国南方电网科技项目(GDKJXM20180152);国家自然科学基金(11704040;11874096)。

摘　　要：提出了一种新颖高效的、超高分辨率的反卷积声源识别方法,即分裂增广拉格朗日收缩(SALSA)反卷积声源识别算法。该方法利用主要声源通常具有的稀疏特性和求解大规模稀疏恢复问题的交替方向思想,在波束形成反卷积数学模型中引入一个和源强等价的分裂变量,进而建立了增广拉格朗日变量分裂声源识别数学模型,并采用SALSA来交替迭代求解该分裂模型获得声源强度。仿真和试验结果表明,该方法与经典的反卷积声源成像方法(DAMAS)相比,源强量化能力相当,还拥有更优的收敛性,在整个分析频率范围内都拥有超高的分辨率,迭代计算速度快数十倍。Here,a novel and efficient deconvolution algorithm with ultra-high resolution for sound source identification was proposed,it was called the split augmented Lagrange contraction deconvolution algorithm(SALCDA).In this method,the sparsity that the main sound source usually has and the idea of alternating directions for solving large-scale sparse recovery problems were used.Then a split variable equivalent to source strength was introduced into the mathematical model of beam forming deconvolution to establish the mathematical model of augmented Lagrange variable splitting for sound source identification.SALCDA was used to solve alternately and iteratively the split model,and gain the source strength.The results of simulation and tests showed that the source strength quantification ability of the proposed method is comparable to that of the classical deconvolution approach for mapping of acoustic sources(DAMAS);it has better convergence,ultra-high resolution within the whole analysis frequency range,and its iterative calculation speed is tens of times faster.

关 键 词：声源识别 稀疏约束反卷积 分裂增广拉格朗日收缩(SALSA) 

分 类 号：TB532[理学—物理]