An offline system for handwritten signature recognition

Automatic online and offline signature recognition and verification is becoming ubiquitous in person identification and authentication problems, in various domains requiring different levels of security. There has recently been an increasing interest in developing such systems, with several views on which are the best discriminator features. This paper presents a new offline signature verification system, which considers a new combination of previously used features and introduces two new distance-based ones. A new feature grouping is presented. We have experimented with two classification methods and two feature selection techniques. The best performance so far was obtained with the Naiumlve Bayes classifier on the reduced feature set (through feature selection).