基于新的变步长模型的LMS自适应滤波算法

摘要

It is difficult for the traditional LMS adaptive filtering algorithms to solve the conflict between convergence speed and steady state error. In order to deal with the problem, a novel nonlinear function relationship between the variable step size factor and the error signal is established, whieh is based on sample function. The performance of the algonthm is improved by functianal relationship through adjusting the exponential factor and other parameters. When the error signal is large at the initial phase of this algonthm, the step size factor can be increased automatically. When the error signal is near to zero, the step size has the property of slight change The experunental result shows tbat the new algorithm has quick convergence speed and small steady-state error, and can solve the ahove conflict effectively. Especially, the convergence speed is increased significantly, and is easy to implement the real-time function.%为了解决传统的变步长LMS自适应滤波算法不能有效处理既要求收敛速度快又要求稳态误差小的矛盾,提出了步长因子与误差信号之间的一种新的基于抽样函数非线性模型.该模型在误差信号上增加了指数因子,在仿真实验中,通过调节该指数因子及其他参数,使得算法在初始阶段误差较大时,步长因子能够自动增大,并在误差信号接近零处,有缓慢变化的特性.实验结果表明,提出的算法有很快的收敛速度和较小的稳态误差,可以有效解决上述矛盾,尤其是大幅度提高了算法的收敛速度,便于实时实现.