Artificial Neural Network based Writer-Independent Offline Signature Verification

• Published in: 2022 3rd International Conference on Intelligent Engineering and Management (ICIEM) pp. 704 - 707
• Main Authors: Kumar, Ashok, Bhatia, Karamjit
• Format: Conference Proceeding
• Language: English
• Published: IEEE 27.04.2022
• Subjects: Artificial neural networks; Artificial-Neural-Network; Handwriting recognition; Local-Binary-Pattern based Features; Offline-Signature Verification; Testing; Training; Writer-Dependent Approach; Writer-Independent Approach
• DOI: 10.1109/ICIEM54221.2022.9853079

Since two or more person's signatures may appear to be identical, but a person's signature may differ reliant on the state, so, signature verification is a problematic research subject. The goal of this study is to see how well an Artificial-Neural-Network and a Local-Binary-Pattern feature set work together to create a Writer-Independent Offline-Signature verification system. The performance of system is evaluated using two datasets of signature, each with 260 and 100 writers. Authentic signatures of a person, as well as skilled-forgery, nonskilled-forgery, and random-forgery signs, are used to test the performance of the developed system, and authentic signatures, as well as skilled-forgery, nonskilled-forgery, and random-forgery signs, are taken into account in the development of the desired system. In this study, a false-acceptance-rate of 23.00 percent, 11.00 percent, and 0.00 percent was obtained for skilled-forgery signs, nonskilled-forgery signs, and random-forgery signs, respectively, while a false-rejection-rate of 0.00 percent was obtained for 15 reference signatures using a database of 260 writer's signatures.