一种新的去相关变步长LMS算法及其应用  

New uncorrelated variable step LMS algorithm and its application

作　　者：金杰[1] 李东辉[1] 徐银浩[1] 

机构地区：[1]天津大学电子信息工程学院,天津300072

出　　处：《计算机工程与应用》2008年第19期57-58,79,共3页Computer Engineering and Applications

基　　金：国家自然科学基金(the National Natural Science Foundation of China under Grant No.60777014)

摘　　要：在一种变步长算法基础上,从语音信号相关性的角度出发,提出了一种新的去相关变步长LMS算法(DCL—NLMS)。该算法结构简单,收敛速度快,稳态失调小,计算量与NLMS算法相当。仿真结果表明,该算法在处理强相关性信号时,不仅收敛速度明显快于其余算法,而且稳态失调特性也有很大优势。Based on the variable step NLMS arithmetic,this paper proposes a new uncorrelated variable step arithmetic（DCL- NLMS）by employing auto-correlation of the input speech signal.The proposed algorithm has simple structure,fast convergence speed,but small stability maladjustment with the same computational complexity as NLMS.Simulation results show not only its convergence is faster than that of the others,but also the stable maladjustment of it is better too.

关 键 词：自适应回波抵消 变步长 相关性 NLMS算法 

分 类 号：TP273[自动化与计算机技术—检测技术与自动化装置]