机构地区: 南京大学计算机科学与技术系计算机软件新技术国家重点实验室
出 处: 《计算机科学》 2005年第5期181-184,共4页
摘 要: 手写签名验证作为一种有效的生物身份认证技术,具有广泛的应用前景,但由于签名易变化的特点,其性能还不够理想。最近神经网络越来越多地在模式识别领域使用并取得了很好的效果,但在签名验证中尚不多见。本文提出并实现了一个基于BP神经网络的动态手写签名验证原型,并通过实验对网络结构进行了分析。通过对从10人收集的190个本人签名和371个伪造签名评价EER=2.16%,还是比较满意的。 Owing to traditional signatures being widely and maturely used in finance or government department hand- written signature verification(HSV),as an effective and efficient kind of biometric based authentication schemes, has a good perspective. But its performace is not good enough because of signatures′easy change. Recently,neural network is coming to used in Pattern Recognition more and more and has archieve good effects. But there are a few of research- es on application of neural network in HSV. This aticale has apposed and implemented a prototype of dynamic HSV based on Back-propagation neural network and has analyzed network structure from experiments. Through 190 own signatures and 371 forged signatures from 10 persons,equal error rate 2.16% are archived.