An Off-Line Signature Verification Method based on the Questioned Document Expert’s Approach and a Neural Network Classifier CESAR SANTOS 1 EDSON J. R. JUSTINO 1 FLÁVIO BORTOLOZZI 1 ROBERT SABOURIN 2 1 PUCPR - Pontifícia Universidade Católica do Paraná, Rua Imaculada Conceição, 1155, Curitiba, PR, Brazil cesar.roberto@pucpr.br {justino, fborto}@ppgia.pucpr.br 2 ÉTS - École de Technologie Supérieure, 1100, rue Notre-Dame Ouest, Montréal, Québec, Canada Robert.Sabourin@etsmtl.ca Abstract In an off-line signature verification method based on personal models, an important issue is the number of genuine samples required to train the writer’s model. In a real application, we are usually quite limited in the number of samples we can use for training (4 to 6). Classifiers like the Neural Network [5], the Hidden Markov Model [2] and the Support Vector Machine [1] need a substantial number of samples to produce a robust model in the training phase. This paper reports on a global method based on only two classes of models, the genuine signature and the forgery. The main objective of this method is to reduce the number of signature samples required by each writer in the training phase. For this purpose, a set of graphometric features and a neural network (NN) classifier are used. Keywords: Signature verification, Expert’s classifier, Neural network. 1. Introduction Usually, two different pattern classes make up an off- line signature verification method in training and verification tasks (W 1 and W 2 ) based on personal models. W 1 represents a genuine signature set for a specific writer and W 2 represents a forged signature set. In the latter case, the set of forgeries is divided into three different types (random, simple and simulated) [2,8]. The random forgery is usually a genuine signature sample belonging to a different writer, one who is not necessarily enroled in the signature verification system. The simple forgery is a signature sample with the same shape as the genuine writer’s signature, while the simulated forgery is a reasonable imitation of the genuine signature model. The training phase uses a set of genuine signature samples (W 1 ) to produce a robust personal model. Usually, a meaningful number of samples capable of representing personal variability make up this set (see Fig. 1). The verification phase uses a personal model to discriminate among writers and among all types of forgeries (W 2 ). Figure 1. The signature models area for different authors in an off-line signature verification method based on personal models Proceedings of the 9th Int’l Workshop on Frontiers in Handwriting Recognition (IWFHR-9 2004) 0-7695-2187-8/04 $20.00 © 2004 IEEE