BP神经网络分类器在签名认证系统中的应用

Application of BP Neural Net Classifier in the Signature Verification System

介绍签名认证系统中基于统计特征向量的BP神经网络分类器的设计,改进BP学习算法,加快训练的收敛速度。利用构造的伪造签名样本和真实签名样本进行训练确定神经网络各层节点的权值,妥善地解决由于人为的伪造样本缺乏,阻碍神经网络分类器的实际应用的问题。试验结果表明,所设计的神经网络分类器在签名认证系统中取得了较好的认证结果。

It introduces the design of BP neural net classifier basing on statistic feature vectors in signature verification system, and improves the learning algorithm of it to quicken the convergent velocity of train. It fixes on the weights of the net using both tectonic forgery signatures and reality signatures, and solves the problem well done that the lack of forgery signatures by people baffles the application of BP neural net classifier. The result of experimentation indicates the designed BP neural net classifier gains a satisfying rate of verification in signature verification system.