基于神经网络的特征结构提取

Neural-network Based for Eigenstructure Extraction

本文在详细讨论特征结构提取问题的能量函数表示的基础上，通过对该能量函数表示的适当变换和整理，高阶关联非线性连续时间神经网络被引入到特征结构的提取问题中。这种方法直观、明了，它将所要提取的特征结构与网络稳定时的输出直接对应起来，可对特征结构进行自适应跟踪、估计，这为特征结构的提取提出了一个新的方法。

In the paper,The expression of the cost function of the eigenstructure extraction problem is discussed in detail.Through a proper transformation of the expression of the cost funtion,high order neural network is introduced into the eigenstructure extraction problem.According to the method, the eigenstructure to be extracted is connected with the output of the network directly when the network is stable.The method is simple and direct,and can be used to track the eigenstructure adaptively.It opens a new area for the extraction of the eigenstructure.