基于滑动窗口的动态手写签名局部相关性研究

Study on Local Correlation of On-line Handwriting Signature Based on Sliding Window

为了解决直接序列匹配中签名序列存在随机波动和时间轴方向非均匀伸缩,导致相关分析给出的匹配度不高的问题,提出了采用滑动窗口对真实签名进行局部相关性分析的方法.将手写签名按压力划分成若干笔段,研究笔段匹配算法;对分段后的数据序列用滑动窗口算法进行局部相关分析.算例显示,对不同的签名个体而言,总有一些笔段的相关性极高,最小相关系数都达到0.8甚至0.9以上,这些笔段,正是签名者的稳定的签名特征,无论是相关分析法还是特征矢量分类法,滑动窗口局部相关分析都是一种有效的算法.

In direct signature matching verification, data fluctuating at random and time wrapping nonlinearly will depress verification veracity of correlation analysis. Based on sliding window local correlation analysis of on-line handwriting signature is presented. Handwriting signature data sequence is segmented according to signature press; then the segments matching algorithm is studied. The process of the local correlation of signature based on sliding window is introduced. Examples reveal that there are always some ＇segments that have very high correlation coefficient, minimum value up to 0.8 to 0.9, in one＇s signatures and those are just the steady signature characters of him. Based on sliding window local correlation analysis is an effective method for handwritten signature verification not only for correlation analysis verification but also for feature vector verification.