论文部分内容阅读
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。
Based on a detailed discussion of the energy function representation of eigen-structure extraction problems, this paper introduced the high-order associated nonlinear continuous-time neural network into the extraction of eigen-structure by proper transformation and arrangement of the energy function representation. This method is intuitive and clear. It directly correlates the feature structure to be extracted with the output of the network when the network is stable, and can adaptively track and estimate the feature structure. This presents a new method for extracting the feature structure.