Issues pertaining to optimal performance of an on-line autoregressive model based signature verification system
N. Mohankrishnan, Mark Paulik, M. Khalil · 2002
The factors determining the optimality of a technique for signature authentication have been examined. These are: procedures for validation and preprocessing of the data, selection of the parameters of the model, and application of higher level decision strategies to enhance performance. The best results were obtained using a sequential strategy requiring the test signature to be repeated a maximum of two times in the event of a defined uncertainty. The false acceptance and false rejection error rates averaged about 0% and 4% respectively using a database of 1600 signatures.>