α稳定噪声环境下ⅡR自适应滤波递归整体最小P-范数算法  

Recursive total least l_p-norm algorithm for adaptive ⅡR filtering in α stable noise environments

作　　者：张斌[1,2] 冯大政[1] 刘建强[1] 

机构地区：[1]西安电子科技大学雷达信号处理重点实验室,陕西西安710071 [2]空军工程大学电讯工程学院,陕西西安710077

出　　处：《西安电子科技大学学报》2009年第6期1015-1020,共6页Journal of Xidian University

基　　金：国家自然科学基金资助(60672128)

摘　　要：当无限脉冲响应(ⅡR)系统输入和输出信号被α稳定噪声干扰时,传统的最小平均P-范数(LMP)算法的解会出现较大偏差,而整体最小平均P-范数(TLMP)算法存在收敛速度慢的问题.为此提出一种适用于自适应ⅡR滤波的递归整体最小P-范数(ⅡR_RTLP)算法,首先整体考虑输入和输出信号受α稳定噪声干扰的影响,使得基于P-范数的误差期望值达到最小;然后采用矩阵求逆引理和幂迭代法递归更新自适应滤波器的系数,使其可跟踪时变系统,并提高算法收敛速度.仿真结果表明,ⅡR_RTLP算法比TLMP算法有较小的系统估计误差和较快的收敛速度.When both the input and the output of a linear system are corrupted by a stable noises, the classical least mean lp-norm （LMP） algorithms usually provide a biased solution and the total least mean lp-norm （TLMP） algorithms suffer from slow convergence. The aim of this paper is to develop a recursive total least lp-norm （IIR_ RTLP） algorithm for adaptive IIR filtering with noisy data. The proposed IIR RTLP algorithm makes the expectation of lp-norm of the error be minimized when both the input and the output are corrupted by a stable noises. In ordor to trace the time-raring system and increase the speed of convergence, the IIR RTLP algorithm recursively updates the adaptive filter coefficients on the basis of the matrix inversion lemma and the power iteration. Simulation results show that the IIR RTLP algorithm can lead to faster convergence and a smaller system error than the existing TLMP algorithms for adaptive IIR filtering.

关 键 词：稳定噪声 自适应滤波 ⅡR系统 递归整体最小P-范数 幂迭代 

分 类 号：TN911.72[电子电信—通信与信息系统]