![]() | Only 14 pages are availabe for public view |
Abstract An important issue in biometrics is identity verification using the authenticity evaluation of a handwritten signature. They are universally uncontroversial for verification objectives, such as authenticating official papers and commercial enterprise contracts. There are several approaches for the verification of signatures using signing dynamics process. The thesis introduces two models for online signature verification, one of them based on neutrosophic rule-based verification system (NRVS) and Genetic NRVS (GNRVS) models. In NRVS, every logical variable is described by three values, truth, indetermincy, and falsity which are determined by neutrosophic membership functions. By these degrees, the proposed model will capture the dynamic global and local features. Besides, NRVS with the three membership functions can deal with all features without a need to feature selection method. It improves the complexity of the model. In GNRVS model, the neutrosophic rules are refined by a Genetic Algorithm. The performance of the proposed system is tested on the MCYT-Signature-100 dataset. The experimental results show that the accuracy, average error rate, false acceptance rate, and false rejection rate have improved when we compare the results with the fuzzy rule-based verification system. The other model is introduced here in this thesis for verifying handwritten dynamic/online signatures, based on deep learning to verify online signatures. It includes more experimental results from one more dataset. In this model, we used coefficients of Legendre polynomials as features with constant length to model signers’ signatures, deep feedforward neural network as the classifier with stochastic gradient descent with momentum as the deep learning algorithm. Principal Component Analysis (PCA) is used to speed up the algorithm by reducing input-dimentionality. Better experimental results are achieved regarding Equal Error Rate reduction, Logarithmic loss reduction and accuracy magnification on MCYT-Signature-100 and SigComp2011 databases in comparison with state-of-the-art methods. |