摘要
提出一种新的非高斯信号激励的非最小相位ＭＡ模型参数辨识方法。该方法基于三阶矩和自适应算法，并利用ＦＫＡ算法极小化输出信号的三阶矩估计方差。计算机模拟结果表明，木文方法的收敛速度比最大梯度下降法（ＮＰＧ）快，与ＮＰＯ算法同数量级，但每次递归的运算量比ＮＰＯ算法少，占内存量也小。
There now exist two MA system methods for identifying non - Gaussian noise(radar signal is special kind of non - Gaussian noise): NPG (steepest descent algorithm) and NPO (Nonminimum phase MA identification with ORIVD[2] and 3- rd order cumulants). We now present a novel MA system method that is better than NPG and NPO.Our method is based on third order cumulants and employs FKA (fast Kalman algorithm ). FKA is a type of recursion least squares algorithm. Eqs. (17a) through (17f) and Eqs. (18a) through (18c) are the iteration equations needed by our method.In Table 1, NPG and NPO are compared with our FKA method. The number of multiplications needed is lowest for FKA. Thus, both operations per iteration and memory space required are less for FKA.In Figs. 2 and 3, simulation results of NPG and our FKA are compared. They both show that convergence rate for FKA is higher.
出处
《西北工业大学学报》
EI
CAS
CSCD
北大核心
1996年第2期290-293,共4页
Journal of Northwestern Polytechnical University
基金
航空科学基金
