Abstract: The colored noise of non-linear systems in stable distribution environments is modeled with Polynomial Autoregression（PAR） model.Parameters are estimated by Expanding Iteratively Reweighted Least P-norm（EIRLP） arithmetic.The properties of Fractional Lower Order Covariance（FLOC） spectrum are given and proved in the dissertation and the frequency domain characteristics of non-Gaussian PAR distribution with infinite variance are analyzed.Theoretical analysis and simulation results show that EIRLP is a robust arithmetic of parameter estimation.By estimating the sinusoidal signal under the stable colored noise environment，the results show that the Fractional Lower Order Covariance（FLOC） spectrum has preferable performance of spectrum estimation.

Key words: stable distribution, non-linear system, Polynomial Autoregression（PAR）, parameter estimation, Fractional Lower Order Covariance（FLOC） spectrum

摘要： 用多项式自回归模型对非线性系统中稳定有色噪声建模，利用扩展的迭代重加权最小[p]范数算法进行模型参数估计。系统研究了分数低阶协方差谱的性质，并对无限方差非高斯多项式自回归有色噪声进行频域特性分析。理论分析和仿真实验表明，EIRLP算法是在高斯和分数低阶稳定分布噪声条件下具有良好韧性的参数估计方法。仿真通过对稳定有色噪声条件下的正弦信号进行谱估计，结果表明，分数低阶协方差谱具有良好韧性的谱估计性能。

关键词: 稳定分布, 非线性系统, 多项式自回归, 参数估计, 分数低阶协方差谱