改进的基于能量的声源定位算法  

Improved solution for energy-based source localization

作　　者：刘艳芳[1] 张会芝[1] 王鲜芳[1] 

机构地区：[1]河南师范大学计算机与信息工程学院,河南新乡453007

出　　处：《计算机工程与设计》2015年第7期1976-1980,共5页Computer Engineering and Design

基　　金：国家自然科学基金项目(61173071);河南省教育厅科学技术研究重点基金项目(13A520534;13A520522)

摘　　要：为提高基于能量的两步加权最小二乘声源定位算法(two-step WLS)的定位精度,提出一种改进算法。通过引入广义矩阵,将有噪数据中的噪声项分离出来作为约束条件,将加权最小二乘问题转化为带约束条件的最小化代价函数的求解问题,采用拉格朗日乘子法求得代数解。仿真结果表明,该算法在噪声较小时,逼近CRLB(Cramer-Rao lower bound)下限;高噪声时,相对two-step WLS算法,提高了定位精度,提高了突然偏离CRLB的噪声阈值,是一种高效可行的定位算法。To reduce the errors of the energy-based two-step weighted least square algorithm（two-step WLS）of acoustic source localization,an improved algorithm was presented.Through the introduction of augmented matrix,the noise term was separated from noise data as its constraint condition,and the weighted least squares problem was converted to the minimum cost function with constraints.The Lagrange multiplier method was used to get the algebraic solution.Simulation results show that,the proposed algorithm reaches the CRLB limit when the noise is not too large.For relative large noise,the proposed algorithm provides much better accuracy than the two-step WLS algorithm,and improves the noise threshold that suddenly deviates from CRLB.The algorithm is an efficient location method.

关 键 词：声源定位 能量 蒙特卡罗下界 闭合表达解 非负约束 

分 类 号：TP391.9[自动化与计算机技术—计算机应用技术]