Online Handwritten Signature Verification System Based on Neural Network Classification

This paper discusses the biometric personal characteristics appliance as an identifying information for granting remote access to information resources. The authors investigated the possibility to apply handwriting dynamics as a characteristic that can be changed in case of its compromise. They proposed their own method of digital data pre-processing based on an algorithm of viewing a straight line as a curved one. The authors analyzed the proposed algorithm for the possibility of applying it to set an online verification. They received positive results. The proposed algorithm allows us to form a feature vector of an info system user of the given dimensionality. The authors suggest using artificial neural network as a classifier with different the most well-known algorithms of their training. The authors propose to use the statistical processing of training set samples with the correlation analysis method of the data, which enters the neural network.