Offline Signature Recognition via Convolutional Neural Network and Multiple Classifiers

One of the most important processes used by companies to safeguard security information and prevent it from unauthorized access or penetration is signature process. As businesses individuals move into digital age, a computerized system that can discern between genuine faked signatures crucial for protecting people's authorization determining what permissions they have. In this paper, we Pre-Trained CNN extracts features forged signatures, three widely classification algorithms, SVM (Support Vector Machine), NB (Naive Bayes) KNN (k-nearest neighbors), these algorithms are compared calculate run time, error, loss, accuracy test-set consist images (genuine forgery). Three classifiers have been applied using (UTSig) dataset; where loss were calculated each classifier in verification phase, results showed got best (76.21), while time (0.13) result among other classifiers, therefore terms our measures.