应用于功率倒置阵的改进最小均方算法

针对功率倒置阵列采用最小均方(LMS)算法迭代求取权值时,收敛速度慢、迭代步长不能随输入信号功率的变化,自动进行调整的弊端,提出采用一种基于指数函数的归一化变步长LMS算法实现功率倒置阵列的方法。分析步长调整原理,通过建立步长和误差之间的非线性函数关系,结合归一化思想,使算法在保证精度的同时具有较快的收敛速度。基于四阵元Y形平面阵,对该算法、LMS算法和归一化LMS算法进行仿真对比,其结果表明,改进算法可以提高功率倒置阵的收敛速度,增大算法的动态输入范围。

Improved least mean square algorithm for power-inversion array

The convergence speed is low and the iterative step size can not be adjusted automatically with the input power change in the power-inversion array when using least mean square(LMS)algorithm.A power inversion array using a variable step size normalized least mean square algorithm based on exponential function was proposed.The principle of step size adjustment was analyzed.The algorithm showed higher convergence speed under the condition of the guaranteed precision by establishing a non-linear functional model between the step size and error,combined with the normalized idea.Based on four element Y shaped planar array,the simulation results show that the ameliorative algorithm outperforms the original LMS algorithm and normalized LMS algorithm in rate of convergence,and also enlarges the dynamic range of input.