摘要
针对传统子空间跟踪算法正交性和稳定性差的问题,基于数据投影算法(DPM),提出了一种步长为对角阵的正交DPM算法。算法运算复杂度较低,能提供标准正交的子空间,运行时无累积性误差;采用"自适应"而非固定搜索步长,能更好地匹配子空间的动态收敛速度,可进一步提高收敛速率并具有更佳的跟踪性能。仿真结果证明了新算法的有效性和正确性。
To solve the bad performance in orthonormality and robustness for the conventional subspace tracking methods,an algorithm based on data projection method（DPM） was proposed in this paper in which a diagonal-matrix step-size was adopted.Besides the lower computational complexity,the method could provide orthonormal subspace without round-off error accumulation.Meanwhile,using the ＂adaptive＂ but not the same step-size could further improve the convergence speed and has a better tracking performance.The effectiveness and efficiency of the proposed algorithm was verified by the simulation results.
出处
《上海航天》
2010年第3期18-21,64,共5页
Aerospace Shanghai
关键词
子空间跟踪
数据投影算法
对角阵步长
Subspace tracking
Data projection method
Diagonal-matrix step-size
