Impact of Forgeries on the Performance of an On- Signature Verification System

The effect of forgeries on the performance of an autoregressive model based signature authentication system is investigated. The forgeries were simulated by making all writers transcribe the same fictitious signature. While there is a degradation in overall accuracy, the authors believe that the nature of the task set for the writers in creating the database is at least partially responsible. A thorough discussion is presented of the lessons learnt from an experimental study.