基于降低运算复杂度的子空间跟踪算法稳健性分析  

Stability and Robustness Analysis for Subspace Tracking Based on Reducing Computational Complexity

作　　者：钱林杰[1,2] 程翥[1] 石斌斌[1] 万建伟[1] 

机构地区：[1]国防科技大学电子科学与工程学院,湖南长沙410073 [2]重庆通信学院,重庆400030

出　　处：《国防科技大学学报》2010年第3期75-81,共7页Journal of National University of Defense Technology

基　　金：国家部委资助项目(41901140401)

摘　　要：FDPM和FOOJA是目前两种最为有效的适用于主、次子空间跟踪的方法,属于低复杂度算法。进一步降低算法运算复杂度,对保证算法的实时性具有非常重要的意义。以降低算法运算复杂度为背景,通过对FDPM、FOOJA算法的分析,指出存在两种简化运算量的FDPM1stcol、FOOJA1stcol方法。在有限精度运算条件下,对四种方法的稳定性、数值鲁棒性进行了深入分析和讨论。通过实验仿真,对分析的结论进行了验证。FDPM and FOOJA,which belong to the low complexity class,are the most efficient algorithms for principal or minor subspace tracking. Reducing computational cost is crucial for guaranteeing the real time implementation. Under lower computational complexity background,it was found that there exist two other simplified methods （FDPM1stcol and FOOJA1stcol） through analysis of FDPM and FOOJA. Under the finite word length condition,the stability and numerical robustness were analyzed and compared with the former two algorithms. Simulation results verified our conclusion.

关 键 词：子空间跟踪 运算复杂度 鲁棒性 稳定性 标准正交 

分 类 号：TN911.72[电子电信—通信与信息系统]