基于神经网络和证据理论的离线签名识别  

Off-line Signature Recognition Based on Neural Networks and Evidence Theory

作　　者：陈刚[1] 李弼程[1] 林琛[1] 

机构地区：[1]郑州解放军信息工程大学信息科学系,450002

出　　处：《微计算机信息》2005年第11S期161-163,共3页Control & Automation

基　　金：河南省教育厅基金(编号:sp200303099)资助项目

摘　　要：论文提出了一种有效的基于神经网络和证据理论的离线签名识别方法。利用从签名图象中提取的三种特征分别构造BP神经网络分类器对签名图象进行初步识别,然后将各分类器的输出作为证据,利用证据理论融合来自不同分类器的输出得到最终识别结果。实验表明,该识别方法能有效提高离线签名的识别率。An effective method based on neural networks and evidence theory for off-line signature recognition is proposed in the paper. Firstly, three sets of features are extracted from a signature image, and then these features are fed to three BP neural networks classifiers for elementary recognition. Finally, recognition result comes out by fusing the previous recognition results from different classifiers with evidence theory. Experimental results show that the proposed method is effective.

关 键 词：离线签名识别 神经网络 证据理论 融合 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]