摘要
针对LMS算法收敛性能差的缺点,提出一种改进LMS算法即NLMSISA。首先从理论上分析并介绍了LMS算法及其改进算法,并将改进算法—变符号函数法(ISA)和归一化变步长(NLMS)算法有机结合,提出了该NLMSISA算法,理论上该改进算法具有物理实现难度低、收敛速度快、且能收敛到更小且稳定的均方误差(MSE)的特点。在MATLAB下,通过采用LMS、LMSISA和NLMSISA三种算法对FIR和IIR两种结构的系统分别进行了辨识仿真研究,结合三种算法的均方误差收敛曲线分析,验证了提出的改进LMS算法在敛速度、稳态误差上都明显优于现有几种算法。该算法对于实际工程应用的自适应滤波问题提供了一个较好的算法。
In this paper, a new adaptive filtering algorithm with fast and variable step size is proposed. On the one hand, theoretical analysis shows that in different channels, the proposed algorithm accomplishes faster convergence and smaller MSE than the LMS and LMSISA. On the other hand , three algorithms (LMS, LMSISA and NLMSISA) are adopted to identify two structure (FIR and IIR ), and the computer simulations demonstrate the characteristics of proposed algorithms. The algorithm is suitable for the adaptive filtering in engineering.
出处
《计算机仿真》
CSCD
2007年第6期89-91,238,共4页
Computer Simulation
关键词
可变步长算法
系统辨识
自适应
滤波算法
Variable step size algorithm
System identification
Adaptive
Filtering algorithm