改进的变误差宽度变阶数LMS算法
2017-02-16马令坤吴波毛红艳
马令坤+吴波+毛红艳



摘 要: 针对自适应滤波器阶数失配问题,提出一种改进的变误差宽度分数阶变阶数LMS算法。该算法中误差宽度函数参数选择不受噪声先验知识的限制,并给出参数选择的依据。对提出的算法分别在高噪声、低噪声以及变化的噪声环境下进行仿真分析,仿真结果表明,该算法在未知系统噪声大小或大小变化的噪声环境能够应用,并且具有良好的性能,尤其在高噪声环境下能够获得较快的阶数收敛速度和较小的稳态误差,因此该算法的应用场合更加广泛。
关键词: 自适应滤波器; 模型失配; 分数阶变阶数; 变误差宽度
中图分类号: TN911?34 文献标识码: A 文章编号: 1004?373X(2017)01?0041?04
Abstract: In order to solve the order mismatch problem of the adaptive filter, an improved variable error width variable fractional tap?length LMS algorithm is proposed. The parameter selection of the error width function isn′t limited by the noise prior knowledge in the algorithm, and its foundation is given. The simulation analysis are performed for the algorithm under the environments of low noise, high noise and changing noise. The simulation results show that the algorithm can be applied to the environments of the unknown system noise or changing noise, has good performance, can obtain fast order convergence rate and small steady?state error in high noise environment, and its applications are extensive.
Keywords: adaptive filter; model mismatch; variable fractional tap?length; variable error width
0 引 言
在自适应滤波应用中,大多数算法一般假设滤波器阶数(或抽头长度)是适配的;但如果滤波器阶数小于实际阶数,将会增大均方误差值;相反,滤波器阶数过大,不仅会带来很大计算量,同时会增大稳态误差。由此可见,滤波器阶数是一个重要参数。
在实际应用中为了能够确定滤波器的最佳阶数,变阶数算法是一类有效的方法,典型算法主要有:分割滤波器LMS算法[1],梯度下降LMS算法[2],分数阶变阶数LMS算法[3?4](Fractional Tap?length Least Mean Square,FTLMS),凸组合变阶数LMS算法[5?6],变迭代参数的变阶数LMS算法[7?8]和变误差宽度变阶数LMS算法[9?10](Variable Error Width Variable Fractional Tap?length LMS Algorithm,VW?FTLMS)等。其中文献[3]FTLMS中滤波器阶数迭代的代价函数采用片段稳态均方误差,滤波器阶数更新依赖于分数阶数变化量的积累,保持了文献[1?2]中算法的优点,该算法简单灵活,是分数阶变阶数类算法的基础。……
