一种有效的基于证据理论的离线签名识别方法  

Effective off-line signature recognition method based on evidence theory

作　　者：陈刚[1] 李弼程[1] 曹闻[2] 刘安斐[1] 

机构地区：[1]信息工程大学信息科学系,河南郑州450002 [2]信息工程大学测绘学院,河南郑州450002

出　　处：《计算机工程与设计》2006年第17期3256-3257,3260,共3页Computer Engineering and Design

摘　　要：提出了一种有效的基于证据理论的离线签名识别方法。从签名图像的3种信息载体中提取出4种特征,利用所提取的4种特征分别构造基于证据理论的分类器对签名图像进行初步识别,将各分类器的输出作为证据,用改进的证据理论合成公式融合不同分类器的输出得到最终识别结果。结果表明:该识别方法能有效地提高离线签名的识别率。An effective method based on evidence theory for off-line signature recognition is brought forward. Firstly, four sets of features of a signature are extracted from its three feature images, and then these features are fed to four k-NN classifiers for elementary recognition. Finally, recognition result is obtained by fusing the previous recognition results from different classifiers with a modified combination rule of evidence theory. Experimental results show that the proposed method is effective.

关 键 词：离线签名识别 证据理论 预处理 特征提取 k-NN分类器 融合 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]