基于笔划匹配的在线手写签名身份认证算法  

On-Line Signature Handwritten Verification Algorithm Based on Stroke Matching

作　　者：栾方军[1] 马驷良[1] 程开东[1] 董险峰[1] 

机构地区：[1]吉林大学数学研究所

出　　处：《沈阳建筑大学学报（自然科学版）》2005年第5期568-572,共5页Journal of Shenyang Jianzhu University：Natural Science

基　　金：信息产业部电子信息发展基金项目(2002546);辽宁省教育厅项目(2004F110)

摘　　要：目的使用计算机把手写签名的图像、笔顺、速度和压力等信息与存储的真实签名样本进行对比,以实时鉴别手写签名真伪.方法笔者给出一种手写签名身份认证的笔划匹配算法,选用签名的速度及压力变化率作为特征向量进行相似度计算,降低签名在坐标和方向上不一致造成的影响,引入加权H′模的距离计算公式来度量输入签名与样本签名的差异.结果经第一届国际手写签名身份认证竞赛样本数据库检验,误拒率为2%,误识率为零.大笔划匹配算法解决了签名由于断笔、干扰等因素造成的笔划数不等的问题;在防止伪造签名的前提下,提取签名的动态特征进行比较,提高了算法的识别率.结论在线手写签名认证具有方便和快捷的特点,适合作为身份认证手段.通过实例检验了算法的有效性.On-line signature handwritten verification is a topic in the field of biometric authentication now. By comparing the picture, stroke, speed and pressure of input signaturs with genuine signatures in database, the computer can distinguish the true or false of a handwritten signature real-timely. Stroke-matching and result judgment are the keys to this problem. This paper presents a stroke-matching algorithm of handwritten signature verification. The algorithm is effectual to be examined by signature data from SVC 2004（First International Signature Verification Competition）. The FRR （False Reject Ratio） is 2.0 % and all forgeries are rejected. The paper also introduces distance calculation formula based on weighting H＇ module to calculate the difference between input signatures and true signatures. On-line signature handwritten verification is very convenient, quick-made, suitable to verification and its effectiveness has been illustrated by practical examples.

关 键 词：在线 手写签名身份认证 模式识别 笔划匹配 识别率 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]