摘要
基于手写体数字的几何结构提取端点及其方向、凸点及其方向、三叉点数和四叉点数等数字特征值,运用单字单网的10个并行BP神经网络进行数字识别,达到很好的识别效果.
Network inputting character is a better method at present. This paper based on free handwritten numbers' geometric structure puts forward a way about picking up numbers' eigenvalue for end and its' direction, protruding point and its' direction, three cross point number, four cross point number and so on, and recognizes number which exerts ten paratactic BP neuron network for single number and single network, therefore it obtains better effect.
出处
《北京工商大学学报（自然科学版）》
CAS
2004年第3期43-45,共3页
Journal of Beijing Technology and Business University:Natural Science Edition
