论文部分内容阅读
提出一种在线签名认证中的特征提取和特征选择的方法.采用一种F-Tablet手写板采集签名数据.该手写板的特点是不仅可记录签名时的字形信息(x,y)序列,还可记录签名时的五维力信息(Fx,Fy,Fz,Mx,My)序列.从每个签名中提取3个等级共188个特征,接着定义特征重要性函数F,然后根据特征的重要性函数F的值对选取的188个特征进行排序,对F设不同的阈值就可完成不同的特征选择.在认证过程中使用SVM算法对选取的特征进行训练,然后用训练所得的模型进行验证.该方法的错误拒绝率为1.2%,错误接受率为3.7%.
A method of feature extraction and feature selection in online signature authentication is proposed.The signature data is collected by using a F-Tablet tablet, which not only can record the sequence of (x, y) glyph information at signatures, but also The five-dimensional force information (Fx, Fy, Fz, Mx, My) sequences can be recorded at the time of signature.A total of 188 features of 3 levels are extracted from each signature, and then the feature importance function F is defined, and then the feature importance The value of the function F is used to sort 188 selected features, and different feature selection can be done by setting different thresholds on F. The selected features are trained by the SVM algorithm in the certification process and then verified by the trained model. This method has a false rejection rate of 1.2% and a false acceptance rate of 3.7%.