Real-time writer verification system using tablet information

在本研究中，我们研究了一种新的在线签名验证方法。它使用了笔的运动。我们采用笔运动的高度和方向变化作为参数来提取签名的个性。片剂获得的所有特征数据都经过预处理。它们由书写持续时间的长度规范化，这是由笔压力的大小决定的。在接下来的实验中，我们使用了三个数据集，分别是笔的运动、签名的形状和书写压力。一般来说，当一个人写他的签名时，时间轴上的数据集会有一些波动。比较了常规线性匹配方法和DP匹配方法的结果。在我们的方法中，通过使用笔的运动、签名的形状和书写压力这三个数据集来独立评估正确的验证率。正确的验证率由接受的真实签名百分比和拒绝的模拟签名百分比平衡的点决定。从24个编写器的实验结果中，得到以下验证率：1)线性匹配方法签名形状：72.4% /书写压力：77.0% /笔移动：90.3%2)DP匹配方法签名形状：73.2% /书写压力：79.9% /笔移动：91.3%虽然传统的签名验证方法大多利用签名形状和书写压力，但通过对比这些参数，验证了DP匹配的笔移动是非常有效的。

In this research, we investigated a new on-line signature verification method. which uses the pen movement. We adopt altitude and direction variation of pen movement as parameter to extract the individuality of signatures. All signature data obtained by the tablet are preprocessed. They are normalized by the length of writing duration, which is decided by magnitude of the pen pressure. In the following experiments, we use three data sets, they are pen movement, shape of signature and writing pressure.In general, there is some fluctuations of the data set in the time-axis when a person writes his signatures. We compare the results of normal linear matching method and the DP matching method. In our method, the correct verification rate is independently evaluated by using each of three data sets which are pen movement, shape of signature and writing pressure. The correct verification rate is decided by the point where the accepted percentage of genuine signatures and the rejected percentage of mimicked signatures are balanced.From experimental results with 24 writers, following verification rates are obtained.1)linear matching methodshape of signature : 72.4% / writing pressure : 77.0% / pen movement : 90.3%2)DP matching methodshape of signature : 73.2% / writing pressure : 79.9% / pen movement : 91.3%Although most of conventional signature verification methods utilize the signature shape and writing pressure, it is verified that the pen movement with DP matching is very effective compared to such parameters.