基于EM算法的宽带信号DOA估计及盲分离  

Broadband DOA Estimation and Blind Source Separation Based on EM Algorithm

作　　者：熊坤来[1] 刘章孟[1] 柳征[1] 姜文利[1] 汪华兴 

机构地区：[1]国防科技大学电子科学与工程学院 [2]解放军61541部队

出　　处：《电子学报》2015年第10期2028-2033,共6页Acta Electronica Sinica

基　　金：国家自然科学基金(No.61302141)

摘　　要：本文提出了一种基于EM算法的宽带信号DOA估计与盲分离方法.首先将宽带混合信号转换到频域,然后综合利用带宽内所有频点信息建立似然函数,在此基础上推导出宽带条件下的EM迭代式,从而实现宽带信号DOA及波形的联合估计.并且本文通过分析EM算法的收敛性,自适应的设定角度搜索空间,提高了算法的运算效率.与传统方法相比,本文方法运用的有效信息更多,因此,其在DOA估计精度及波形恢复性能方面都更有优势.仿真实验表明了该算法的有效性.Based on the expectation-maximization （EM）algorithm,a new method for broadband direction of arrival （DOA） estimation and blind source separation is presented in this paper.First,we establish a likelihood function in the frequency domain by jointly using all frequencies information.Then,based on this function,the E-step and M-step in the frame of broadband EM algo-rithm are derived.Finally,we jointly estimate DOAs and sources through EM iteration compute.Moreover,through analyses the con-vergence of EM algorithm,we set adapted angle search space for DOA estimation,which reduce the computations of our algorithm. Compare to the traditional method,more effective information are applied in our algorithm.Therefore,the proposed algorithm has more advantages for DOA estimation and source separation.Simulation results demonstrate our algorithm is effective.

关 键 词：宽带信号 EM算法 DOA 估计 信号分离 

分 类 号：TN971[电子电信—信号与信息处理]