【摘要】 为了解决最小均方(leastmeansquare,LMS)算法中收敛速度和稳态误差之间的矛盾,在独立假设的条件下,从滤波器系数均方误差的角度,提出了最优步长定理,证明了最优步长和均方误差之间存在一一对应的关系。并以此构造了最优变步长LMS(optimalvariablestep-size,OVS-LMS)模型。推出了最优步长的递推式,讨论了最优初始化相对步长的选取方法。综合以上的分析结果,提出了该模型的实现算法。计算机仿真证明了该算法和OVS-LMS模型的学习曲线是非常相近的,因而该算法在独立假设条件下是最优的变步长LMS算法。更多还原

【Abstract】  An optimal stepsize theorem is presented for the on Least Mean Square (LMS) algorithm based on the mean square error (MSE) of the filter weight assuming independent data. The theorem presents a simple relation between the optimal stepsize and the MSE. An optimal variable stepsize LMS (OVSLMS) model is then developed with a method to initialize the step size. A practical algorithm was then proposed and validated by computer simulations. The results show the convergence speed is very close to the theoretical limit for variable stepsize LMS.更多还原