Abstract: This paper discusses the process of off line signature recognition and verification system using neural networks. The paper consists of two stages 1) Evaluating feature vectors of off line-signature images. 2) Training and Validation using back-propagation artificial neural network. Three layer feed forward MLP classifier have been used for pattern recognition process. The main advantage of feed forward neural network is that they are easy to use, and that they can approximate any input/output mapping with weights and thresholds (biases) of the model. Different stages of the process are explained in the paper by making use of proper algorithms. The overall accuracy achieved is 96.3%.

Keywords: Confusion matrix, Feed forward neural network, Kurtosis, Receiver operating characteristics (ROC).