α稳定分布下的自适应数据块滤波算法  

Adaptive data block filtering algorithms for the α-stable distribution

作　　者：赵知劲[1] 董科海[1] 尚俊娜[1] 孔宪正[1] 

机构地区：[1]杭州电子科技大学通信工程学院

出　　处：《电路与系统学报》2009年第3期28-32,共5页Journal of Circuits and Systems

基　　金：国家自然科学基金项目(60872092);浙江省自然科学基金项目(Y106839)

摘　　要：本文基于分数低阶统计量原理提出了α稳定分布下的自适应数据块LMP和LMAD滤波算法。新算法利用更多的输入信号和误差信号的信息,更准确地估计梯度,调整自适应滤波器权向量增量方向,提高了收敛速度。采用变步长因子改善算法的稳态误差,进一步提高了算法的收敛性能。仿真结果表明,在分数低阶α稳定分布信号条件下,自适应数据块滤波新算法较NLMP算法和NLMAD算法有更好的收敛性能。This paper proposes adaptive data block LMP and LMAD filtering algorithms for the a-stable distribution based on the fractional lower order statistics theory. The convergence rate of the proposed algorithms is largely increased by changing the direction of coefficient increment vector with more precise gradient estimation by using abundant information of the error and input signals. Also, a variable step size is applied to the presented algorithms, which further improves the convergence performance. Simulations results indicate that data block algorithms achieve better convergence capability comparing to the NLMP and NLMAD algorithms under the α-stable distribution environment.

关 键 词：Α稳定分布 分数低阶统计量 数据块 LMP算法 

分 类 号：TN911.7[电子电信—通信与信息系统]