改进的盖尔圆源数目估计方法  

A modified Gerschgorin disks estimation method for source number estimation

作　　者：董姝敏 梁国龙 

机构地区：[1]哈尔滨工程大学水声技术国家级重点实验室,黑龙江哈尔滨150001 [2]吉林师范大学信息技术学院,吉林四平136000

出　　处：《哈尔滨工程大学学报》2013年第4期440-444,共5页Journal of Harbin Engineering University

基　　金：国家自然科学基金资助项目(51009042);高等学校博士学科点专项科研基金资助项目(20102304120030);黑龙江省自然科学基金资助项目(E201024)

摘　　要：为提高盖尔圆(Gerschgorin disks estimation)法实现源数目估计的算法性能,提出一种改进的盖尔圆方法(modi-fied Gerschgorin disks estimation,MGDE).MGDE算法主要是利用盖尔圆圆心信息对盖尔圆半径进行独立压缩,使噪声盖尔圆半径压缩速度快于信号盖尔圆半径,从而使噪声盖尔圆尽可能地远离信号盖尔圆,更有利于源数目的准确估计,而且不必人为选择调整因子.仿真验证了MGDE算法在白噪声和色噪声中的源数目估计性能,并与AIC(akaike informationceriterion)、MDL(minimum description length)、及GDE算法进行了对比分析.结果表明:在一定条件下,MGDE算法在白噪声和色噪声中的性能优于GDE算法,而且该算法的小样本源数目能力很好.This paper introduced a modified Gerschgorin radii method called modified Gerschgorin disks estimation(MGDE),in efforts to further improve the performance of Gerschgorin disks estimation(GDE) method for source number estimation.The MGDE method independently minimized Gerschgorin radii by using center information of Gerschgorin disks to make the compression speed of noise Gerschgorin radii faster than that of signal Gerschgorin radii.Also,the noise Gerschgorin disks were kept as remotely away from the signal Gerschgorin disks as possible.Thus,allowing the source number to be determined easily,and no need for the adjustable factor to be chosen subjectively.The computer simulation verified the performance of MGDE algorithm under Gaussian white noise and colored noise,respectively.The MGDE method was then compared with Akaike information criterion(AIC) and minimum description length(MDL) and GDE.The results show that the MGDE algorithm both in Gaussian white noise and colored noise was better than GDE under some constraints,and its performance of estimating the number of sources was very good when the sample number is small.

关 键 词：源数目估计 盖尔圆 色噪声 白噪声 

分 类 号：TN911[电子电信—通信与信息系统]