论文部分内容阅读
提出一种基于支持向量数据描述方法的在线签名身份认证系统.首先,采用双向后向合并DTW算法确定签名中关键点之间的对应关系,然后采用经典DTW度量签名局部中各种细微的差异.文中提出基于差异值均值方差最小原则的特征选择方法.最后,采用支持向量数据描述方法设计分类器.为得到更好的认证效果,采用多层交叉验证和遗传算法寻找最优的分类器参数.在SVC2004数据库上,系统对熟练伪造签名取得4.25%的平均等错误率.
This paper proposes an online signature authentication system based on support vector data description method.Firstly, the bidirectional backwards combined DTW algorithm is used to determine the corresponding relationship between the key points in the signature, and then the subtle differences in the local part of the signed DTW are measured. In this paper, a feature selection method based on the minimum variance of means of mean variance is proposed.Finally, the classifier is designed by using support vector data description method.In order to get better authentication results, multi-layer cross-validation and genetic algorithm are used to find the optimal classifier parameters. On the SVC2004 database, the system achieved a 4.25% average equal error rate for skilled counterfeit signatures.