一种新的变步长自适应最小均方算法

New variable step size least mean square(LMS) algorithm

提出了一种根据滤波器系数梯度差值的自相关来计算步长的新的变步长自适应LMS算法。分析了算法的收敛性能和稳态特性,给出了算法参数选择的原则。由实验验证了该算法具有良好的收敛性能和跟踪特性,特别是在输入信号相关的情况下,该算法显示出比标准LMS算法和其它变步长算法的优异性能。

This paper presented a new adaptive variable step size LMS algorithm, which controls step size according to the correlation of the gradient change of filter coefficients. The convergence performance and steady state characteristic were analyzed. The principle of choice of parameters for the algorithm was introduced. The computer simulations show that the new algorithm has very good convergence ability and tracing ability.Compared with other variable step size algorithm, the algorithm works well specially in correlate input environment.