Online Signature Verification using Time Warp Edit Distance based kernel
Sarojit Auddya, Rajat Kumar Singh, Suresh Sundaram · 2020
In this paper, we propose a new method for dynamic signature verification. We adapt a metric known as Time Warp Edit Distance (TWED) for computing the similarity score between two online traces of signature data samples. We derive a set of secondary features from the online trace of a user's signature and further propose to employ a sequence of Autoregressive coefficients. The TWED algorithm is applied on the set of features obtained from the online traces of signatures. In our approach, we use Support Vector Machine for classification of the samples as either genuine or forgery. We formulate a new kernel function based on the TWED metric in the SVM framework for verification. To demonstrate the efficacy of our proposed algorithm, we perform a set of experiments on the publicly available MCYT-100 database. We present improvements over several state of art works.