Dynamic Alignment Distance Based Online Signature Verification
Ethem Alpaydın · 2004
A self-contained application that verifies the signatures of enrolled users is developed. The verifier uses dissimilarity values based on dynamic alignment distances. Three different threshold calculation techniques are investigated: The first one makes use of the information provided by forged signatures and the other two methods use the information gathered from genuine signatures. The methods are evaluated on two different datasets: One with skilled forgery and the other with random forgery. Best results are obtained by using both genuine and forgery information to calculate the threshold. The lowest reported total error value for the dataset with skilled forgeries is 7.83 percent.