局域动态特征弯曲校正和作用均衡签名认证方法

Signature Verification Based on Characteristic Parts of Signing Signal Through Warp Rectifying and Effect Balancing

利用手写签名动态信息进行身份认证是一种安全性高的个人身份认证方法.由于书写过程稳定性不一致,签名动态信息在时间轴上表现出平移、压缩和舒张等弯曲现象,但某些时间区段内动态信息的稳定性较高,弯曲现象相对微弱,这些特征鲜明的局部区域是签名认证的重要依据.通过设计核变换矩阵选取特征相对稳定的局部动态信息,并根据它们稳定程度的高低来均衡它们在签名认证中的作用,可以避免签名动态信息上不稳定局部对认证结果的负面影响.对挑选的特征局部进行非线性时间弯曲校正,与签名全程动态信息的时间弯曲校正相比较,不仅降低了计算量,而且可以避免校正误差积累.利用校正结果生成核变换矩阵,进而对线性支持向量机实施核化,得到的非线性签名认证器在实验中获得了比较好的认证准确率.

It is with high security to identify personal through signature verification using dynamic signals. As inconsistencies exist in the process of signature signing, the dynamic signals are usually warped to some extent, with contraction, dilation and shift in time. Some parts of the dynamic signals are more stable and with evident features than else. They are the characteristic parts for signature verlfieation. Through picking out these characteristic parts of the signing signals, and balanelng their effects in kernel matrix according to the degree of their stableness, the negative influences on signature verification of the other parts with low stableness can be reduced to the smallest. A nonlinear time warping rectification algorithm is used to deal with the time warping of each characteristic part selected. The kernel matrix with the results of time warping rectification is used to kernelize the linear Support Vector Machine （SVM）, which realizes the aim of verifying signatures non-linearly. The experiment of signature verification demonstrates that the method can obtain a high verifying correct rate.