电子学报 ›› 2017, Vol. 45 ›› Issue (8): 1989-1994.DOI: 10.3969/j.issn.0372-2112.2017.08.026

• 学术论文 • 上一篇    下一篇

一种基于改进PSO的随机最大似然算法

宋华军1, 刘芬1, 陈海华2, 张鹤1   

  1. 1. 中国石油大学(华东)信息与控制工程学院, 山东青岛 266580;
    2. 中国石油大学(华东)计算机与通信工程学院, 山东青岛 266580
  • 收稿日期:2016-07-15 修回日期:2016-09-06 出版日期:2017-08-25
    • 通讯作者:
    • 陈海华
    • 作者简介:
    • 宋华军,男,1978年生,博士,副教授,硕士生导师,毕业于中国科学院长春光学精密机械与物理研究所,美国弗罗里达大学访问学者,主要研究领域为实时目标跟踪算法、高性能硬件系统设计及无线通信.E-mail:huajun.song@upc.edu.cn
    • 基金资助:
    • 国家自然科学基金 (No.61305012)

A Stochastic Maximum Likelihood Algorithm Based on Improved PSO

SONG Hua-jun1, LIU Fen1, CHEN Hai-hua2, ZHANG He1   

  1. 1. College of Information and Control Engineering, China University of Petroleum(East China), Shandong, Qingdao 2665802, China;
    2. College of Computer and Communication Engineering, China University of Petroleum, Shandong, Qingdao 266580, China
  • Received:2016-07-15 Revised:2016-09-06 Online:2017-08-25 Published:2017-08-25
    • Supported by:
    • National Natural Science Foundation of China (No.61305012)

摘要: 随机最大似然算法(Stochastic Maximum Likelihood,SML)具有优越的波达方位(Direction-of-Arrival,DOA)估计性能,但SML解析过程较高的计算复杂度限制了该算法在实际系统中的应用.针对SML计算复杂度高的问题,提出一种低复杂度的粒子群优化算法(Particle Swarm Optimization,PSO),解决了传统PSO算法中粒子数多和迭代次数多的双重缺点.首先,根据天线获得的信号,将旋转不变子空间法(Estimation of Signal Parameters via Rotational Invariance Techniques,ESPRIT)求得的闭式解作为DOA的预估计值,同时计算系统此时的信噪比以及SML在此信噪比下的克拉-美罗界(Cramer-Rao bound,CRB).然后,根据DOA预估计值和当前CRB值在SML最优解的近邻范围内确定较小的初始化空间,并在该空间初始化少量粒子.最后通过设计合适的惯性因子w,使粒子以合理的速度搜索最优解.实验结果表明,改进PSO算法所需的粒子个数和迭代次数大约是传统PSO算法的1/5,降低了SML的解析复杂度,计算时间是传统PSO算法的1/10,因此在收敛速度上也有显著的优势.

关键词: 波达方位估计, 粒子群优化算法, 随机最大似然算法, 计算复杂度

Abstract: The Stochastic Maximum Likelihood (SML) achieves exceptional performance of estimating Direction-of-Arrival (DOA).However,the high computational complexity of analytic method limits SML for further applications in practice.Considering the high computational complexity of SML,we propose a low complexity improved PSO algorithm,which outperforms the traditional PSO approach both in the number of particles and iterations.Based on the signals received by antenna,we firstly obtain the closed solution of Estimation of Signal Parameters via Rotational Invariance Techniques (ESPRIT) to pre-estimate the DOA.In addition,we compute the current Signal Noise Ratio (SNR) of the system as well as the SNR based Cramer-Rao Bound (CRB) of the SML.According to the pre-estimated DOA and current CRB,we then determine a small specific initialized space which is closed to the optimal solution of SML.Besides,we set a few particles in the corresponding search space.Finally,we construct the appropriate inertia factor which lead to an appropriate search speed for particles.Experimental results demonstrate that the number of particles and iteration times required by the improved PSO algorithm is about one-fifth of the traditional PSO algorithm,which greatly reduces the computational complexity of SML,the computation time is one-tenth of the traditional PSO algorithm,thus,the proposed method achieves significant merit of convergence speed.

Key words: direction-of-arrival estimation, particle swarm optimization, stochastic maximum likelihood algorithm, computational complexity

中图分类号: