快速收敛变换域自适应滤波算法  

Fast convergence transform domain adaptive filter algorithm

作　　者：宁小玲[1] 刘忠[1] 冯成旭[1] 罗钦发 

机构地区：[1]海军工程大学电子工程学院指挥自动化系 [2]海军92146部队

出　　处：《电子测量与仪器学报》2011年第3期240-245,共6页Journal of Electronic Measurement and Instrumentation

基　　金：国家高技术研究发展计划(863计划)(编号:2007AA01Z309)资助项目

摘　　要：为了提高LMS算法均衡相关信号时的收敛性能,提出了一种新的快速收敛变换域自适应滤波算法(VS-DCTLMS)。该算法融合了归一化离散余弦变换LMS算法和变步长离散余弦变换LMS算法的优点,具有结构简单,收敛速度快,稳态误差小的特点。仿真结果表明,该算法计算量和变步长归一化LMS(VS-NLMS)算法、解相关LMS(DescorrLMS)等算法相当,但在收敛速度上具有很大优势;收敛性能和最小二乘回归算法(RLS)接近,却克服了RLS计算量大、不利于硬件实现的实际问题,是一种性能优良的可实现方法。A new fast convergence transform domain adaptive filter algorithm（VS-DCTLMS） is proposed in or-der to improve the convergence performance of LMS algorithm for equalizing correlation signal.The new algorithm,which is simple,fast convergence and small steady state error,combines the merits of normalized discrete cosine trans-form LMS algorithm and variable step-size discrete cosine transform LMS algorithm.The simulation results show that computational quantity of the proposed algorithm is comparable to variable step-size normalized least square error（VS-NLMS） and decor relation LMS（DescorrLMS）,but convergence rate of the presented algorithm is superior to VS-NLMS and DescorrLMS on a large extend;the convergence ability of the proposed algorithm is close to recursive least square（RLS） algorithm,but it avoids the practical problems just like great computational quantity and improbable hardware realization which existed in RLS algorithm.This algorithm is a realizable method with good performance.

关 键 词：离散余弦变换 变步长 归一化 收敛速度 均方误差 

分 类 号：TP39[自动化与计算机技术—计算机应用技术]