Handwritten signatures are widely used and well-accepted biometrics for personal authentication. The accuracy of signature verification systems has significantly improved in the last decade, making it possible to rely on machines in particular cases or to support human experts. Yet, based on only few genuine references, signature verification is still a challenging task. The present paper provides a comprehensive comparison of two prominent string matching algorithms that can be readily used for signature verification. Moreover, it evaluates a recent cost model for string matching which turns out to be particularly well suited for the task of signature verification. On three benchmarking data sets, we show that this model outperforms the two standard models for string matching with statistical significance.

中文翻译：

---

基于字符串编辑距离的在线签名验证

手写签名被广泛用于个人认证，并且被广泛接受。在过去的十年中，签名验证系统的准确性得到了显着提高，从而可以在特定情况下依靠机器或为人类专家提供支持。但是，仅基于很少的真实参考，签名验证仍然是一项艰巨的任务。本文提供了两种可以方便用于签名验证的著名字符串匹配算法的全面比较。此外，它还评估了用于字符串匹配的最新成本模型，结果证明该模型特别适合于签名验证任务。在三个基准数据集上，我们显示出该模型优于具有字符串统计意义的两个标准模型。