基于最大峰度准则的非因果AR系统盲辨识

本文结合最大峰度准则和非线性优化理论中的梯度法，设计了一种因果ＡＲ系统的盲辨识算法，并证明了它的全局收敛性，给出了它在平衡点附近的收敛速度，算法通过构造逆滤波器的方法来进行盲辩识，同时通过基于高阶累积量的自 法将逆滤波器的系数逼近ＡＲ系数的参数，由于采用了高阶累积量，处地高斯观测噪声有较好的抑制能力，仿真的结果表明了算法的有效性。

Blind Identification of Noncausual AR System Based on Maximum Kurtosis Criteria

Based on the maximum kurtosis criteria and gradient decent method, this paper designed a new blind identification algorithm for noncausual AR system. Its global convergence is proved and convergence rate calculated. Parameters inthe invease filter approach the true AR coefficients by high-order cumulant (HOS) based self-study algorithm. This identification method can effectively suppress the Gaussian noise because of the utilization of HOS.