基于累积量增强全矢量正交子空间方法的MA建模

提出了一种新的线性代数方法——累积量增强全矢量正交子空间方法(CEAVOS)用于非最小相位非高斯滑动平均(MA)建模，该方法利用组合特性映射的累积量增强，并用全矢量正交于空间法估计MA参数．数值仿真结果表明，CEAVOS的性能优于组合累积量切片法WS和全矢量正交子空间法(AVOS)这2种现有的性能最好的MA参数估计方法，尤其是在估计的偏差上；在低信噪比与短数据的情况下，性能也表现良好．

Identification of MA model based on cumulant enhancement AVOS

A new linear algebraic approach is proposed for identification of a nonminimum phase moving average (MA) model based on third order cumulants from the noisy observations, which is the cumulant enhancement all vector orthogonal subspace method (CEAVOS). This method implements a cumulant enhancement (CE) method based on composite property mappings to obtain better estimation of the cumulants and adopts all vector orthogonal subspace (AVOS) algorithm to estimate the MA parameters. Numerical simulation results show that the performance of CEAVOS is better than that of the two prevailing MA estimation methods: weighted slices (WS) and AVOS, especially with respect to the bias of estimations. The proposed method works well even with shorter data length or at lower SNR.