一种可编程Hamming神经网络特征提取器

A Programmable Hamming Neural Network Feature Extractor

根据 Ham ming神经网络的工作原理提出了一种可用于手写体数字识别的电流型可编程局部结构特征提取电路 .该特征提取器的模板不仅是可编程的 ,它可根据不同的需要来随时更改模板的内容以适应不同的情况 ,而且其特征合并的方式也是可配置的 ,它可根据不同的需要把所需的特征合并成不同的特征类别 .对于该可编程特征的电路模拟以及用单层多晶、双层金属 1.2 μm数字

A current mode programmable local feature extractor for handwritten digit recognition is put forward based on the principle of Hamming neural network.The templates of the extractor is programmable and their contents can be refreshed according to different requirement.The way of feature merging is also reconfigurable so as to merge different features into different feature classes.A prototype chip of the feature extractor has been implemented in single poly,double metals 1 2μm digital CMOS technology and both the simulated and measured results prove that the extractor can realize the feature extraction well.[KH8/9D]