基于皮尔逊相关系数的动态签名验证方法

Dynamic signature verification method based on Pearson correlation coefficient

针对动态签名验证中存在的动态特征长度不等、动态签名验证方法较复杂以及识别率较低等问题,提出了一种基于皮尔逊相关系数的动态签名验证方法。首先通过划分原始特征区域,筛选并计算对应区域内的特征权重和,然后利用皮尔逊相关分析法计算各签名特征间的相关系数;再将皮尔逊相关系数作为新特征,分析真伪签名的皮尔逊相关系数分布情况;最后结合高斯密度函数模型,并通过设置个体判别阈值来进行签名验证。实验结果表明,真签名内的皮尔逊相关系数普遍高于真伪签名间的皮尔逊相关系数,且本方法在SVC和xLongSignDB数据集上均展现了较优的签名验证性能,其中xLongSignDB数据集上的误拒率和误识率分别为2.1%和1.7%。

The dynamic signature verification has problems of the unequal length of dynamic features, complex dynamic signature verification methods, and low recognition rate. To address these issues, a dynamic signature verification method based on correlation coefficient is proposed. First, the feature weight sum in the corresponding region is filtered and calculated by dividing the original feature region. And the correlation coefficients between signature features are calculated by the Pearson correlation analysis method. Secondly, the Pearson correlation coefficient distribution of genuine and simulated signatures is analyzed with the correlation coefficient as a new feature. Finally, the signature is evaluated by combining the Gaussian density function model and setting an individual discrimination threshold. Experimental results show that the Pearson correlation coefficient inside the genuine signature is generally higher than that between the genuine and simulated signatures. This method shows better signature verification performance on SVC and xLongSignDB data sets. The false rejection rate and false acceptance rate on xLongSignDB data sets are 2.1% and 1.7%, respectively.