摘要
基于神经网络对手写体文字多种特征的良好适应性，提出以现有的多种特征分别作为输入，构成多个神经网络识别子系统，将这些子系统有机地结合，以期集成系统具有较高的识别率和很低的误识率。文中给出串联和并联2种集武系统结构，同时分析了2种识别系统的性能，最后通过实验进行了验证。
Relying on the excellent adaptability of the neural network to multiplekinds of features of handwritten characters, the construction of recognition subsystems is suggested, built upon multiple neural network with the current multiplefeatures working as input and the systematic integration of them. Improvements inrecognition and misrecognition rate can thus be achieved. Two integrated systemstructures in the series and parallel connection mode respectively are proposed. Theperformance of these integration systems is analysed. Experimental results showthat the proposed systems are successful.
出处
《兵工学报》
EI
CAS
CSCD
北大核心
1998年第4期339-343,共5页
Acta Armamentarii
基金
国防科工委兵器预研基金!!YJ96095
关键词
神经网络
文字识别
特征提取
neural network,character recognition,feature extraction
