一种α稳定分布下的变抽头长度LMP算法  

Variable Tap-Length LMP Algorithm in α-stable Distribution Noise

作　　者：赵知劲[1,2] 郑晓华[1] 尚俊娜[1,2] 

机构地区：[1]杭州电子科技大学通信工程学院,浙江杭州310018 [2]中国电子科技集团第36研究所通信系统信息控制技术国家级重点实验室,浙江嘉兴314001

出　　处：《电声技术》2010年第11期76-78,共3页Audio Engineering

基　　金：国家自然科学基金项目(60872092)

摘　　要：针对无法事先知道滤波器抽头长度,且系统中存在α稳定分布噪声的情况,提出了一种α稳定分布背景下的变抽头长度自适应算法,采用3个并行的滤波器,比较数据段内其平均P阶误差大小,每K点数据段更新一次抽头长度,数据段内滤波器权值采用LMP算法调整。算法仿真表明,算法在不同特征指数的α稳定分布下都具有很强的韧性,能自适应调整到最优抽头长度,达到最优抽头长度的LMP算法的稳态误差。In many practical application,α-stable distribution can model impulsive random noise more efficiently. For the length of filter taps can not be known in advance and there can be α-stable distribution noise in system, a variable tap-length adaptive algorithm in α-stable distribution noise is proposed. The three parallel filters are used to compare the average error of p-rank and the tap-length is updated every K point data block, and the LMP algorithm is adopted to adjust the weights of filter in the every K point data block. The simulation results indicate that this algorithm is robust in α-stable distribution noise with different characteristics and can adaptively adjust to the optimal tap-length. Meanwhile, it can reach the steady-state error of the optimal tap-length LMP algorithm.

关 键 词：Α稳定分布 变抽头长度 LMP算法 最小分散系数准则 

分 类 号：TN911[电子电信—通信与信息系统]