特小样本动态系统的一种有效的参数估计方法

AN EFFECTIVE METHOD OF PARAMETER ESTIMATION FOR THE DYNAMIC SYSTEM WITH EXTRA SMALL SAMPLES

本文提出了一种特小样本动态系统的极大似然参数估计方法(ESML)。本法是基于预测拟合优势的概念,即对正态分布最佳的拟合是学生分布,而在小样本下其拟合优势是更为显著的。文中推导了简化的目标函数及其优化算法。仿真实例表明了本法的有效性。

This paper describes a maximum likelihood parameter estimation method for dynamic systems with extra small samples (ESML)。The method is based on the concept of prediction fit in which the best fit for normal distribution is student distribution and the goodness of fit is even more conspicuous for small samples。The reduced objective function and the approach of optimization are deduced。Simulation examplesshow good results of this method。