基于多级神经网络结构的手写体汉字识别

A Multi stage Neural Network Architecture for Handwritten Chinese Character Recognition

本文提出了一种用于手写体汉字识别的多级神经网络结构（Ｍｕｌｔｉ－ｓｔａｇｅＮｅｕｒａｌＮｅｔｗｏｒｋＡｒｃｈｉｔｅｃｔｕｒｅ，ＭＮＮＡ）模型。在该模型中，我们将多个神经网络和不同的特征提取方法有机地集成在一起而构成一个完整的模式识别系统。我们讨论了设计ＭＮＮＡ的一般原理，并提出了一个基于多层前馈神经网络的三级结构的手写体汉字识别实验系统。三种不同的特征提取方法被应用于各级子系统之中。对１００个汉字１５０００个样本的实验我们得到了９９．３４％的识别率，０．３６％的拒识率和０．３％的误识率。

In this paper, we propose a Multi stage Neural Network Architecture (MNNA) which integrates several neural networks and various feature extraction approaches into an unique pattern recognition system. General mechanism for designing the MNNA is presented. A three stage fully connected feedforward neural networks system is designed for Handwritten Chinese Character Recognition (HCCR). Different feature extraction methods are employed at each stage. Experiments show that the three stage neural network HCCR system has achieved impressive performance and the preliminary results are very encouraging.