Open AccessSignify: Signature Verification Technique using Convolutional Neural Network
Author(s) -
Alexandra Mae C. Laylo,
Mark Daryl A. Decillo,
Louie Andrew F. Boo,
Jeffrey S. Sarmiento
Publication year - 2019
Publication title -
international journal of recent technology and engineering
Language(s) - English
Resource type - Journals
ISSN - 2277-3878
DOI - 10.35940/ijrte.b1015.078219
Subject(s) - signature (topology) , convolutional neural network , computer science , biometrics , pattern recognition (psychology) , authentication (law) , artificial intelligence , feature extraction , feature (linguistics) , artificial neural network , mathematics , computer security , linguistics , philosophy , geometry
Signature is one of the biometric traits that are being used in person authentication and due to its dominant usage; it became one of the top subjects of forgery. In this study, a signature verification using Convolutional Neural Network (CNN) is proposed. With the use of transfer learning, inception-v3 is mainly used for the feature extraction of data samples and for classification of signatures. The proposed method is assessed on dataset of handwritten signatures gathered from 4 people with 100 signatures each. The testing results determine the threshold value which is 96.43%. Factors that affect the accuracy of the result were also identified.