手写体字符识别的人工神经网络方法的研究  

The Study of Artificial Neural Network for Hand-Written Numeral Recognition

作　　者：于利华[1] 杨福生[1] 刁颐民[1] 

机构地区：[1]清华大学电机系

出　　处：《信号处理》1995年第3期153-161,共9页Journal of Signal Processing

摘　　要：本文总结了生物视觉神经系统从视网膜到初级树皮层的信息处理特点，并指出其与小波交换概念相似处。在此基础上设计了人工神经网络字符识别器，用于数码识别．仿真结果初步显示该模型具有较好的推广能力和分类能力．采用广义能量函数法进一步改善了模型性能。Based on a review of the features of visual information processing of biological system and a comparision of it with wavelet transform , a new ANN architecture and algorithm is devised for the recogition of hand -written numerals . The Generalized Energy Constrained Method is used to improve its performance further . The model is shown to have fairly good generalization and classification capability by computer simulation .

关 键 词：字符识别 文字识别 模式识别 人工神经网络 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]