Model approach to DTW signature verification using error SIGNALS

摘要

Traditional use of Dynamic Time Warping (DTW) methods for signature verification consists of forming a dissimilarity measure between the questioned signature and a set of template signatures. Replacing the template signatures with a hidden signature opens a number of new possibilities for signature analysis. A hidden signature is one that is created based on averaging the measurements of a group or set of signatures. The authors in this paper use the statistical properties of a hidden signature to construct the dissimilarity measure. This approach was tested on the MCYT database using both genuine signatures and skilled forgeries. The results obtained for various combinations of warping spaces and dissimilarity measures show that this method is efficient to verify digitized signatures used for commercial purposes. An appendix and glossary of terms are included at the conclusion of the paper.