Online handwriting verification with safe password and increasing number of features

Abstract

In this Article we present a solution to verify user with safe handwritten password. To achieve the best possible results we use 39 parameters beside some statistical parameters we use primarily time, speed and relation parameter. For the tests a database with 32 users was provided. Each user wrote safe password in one session 12 times. For the FAR (false acceptance rate) test we work with 3 forgeries per user. The best result of 100% success rate for a correct classification, we reached with the Bayes Net classifier. The best result of 3.13% false accepted rate is reached with Bayes Net classifier too.