Handwritten Signature Verification Using CNN with Data Augmentation
Md. Tariqulhasan Fazle Rabbi, Shafizur Rahman, Prokash Biswash, Jinsul Kim, Alamin Sheikh, Aloke Kumar Saha, Mohammad Shorif Uddin · The Journal of Contents Computing · 2019
A signature is a mark or sign which is made by an individual on an instrument or document to signify knowledge, approval, acceptance or obligation. To authenticate writing or a notice of its source and to bind the individual signing which is written by the provisions contained in the document. Signature verification is more important for not only in commercial banks but also with every sector like falsification of documents in numerous financial, legal and other commercial aspects. A signature is an important factor in biometric technique in which it is used to detect forged or genuine signature. This paper concerns offline handwritten signature verification using convolutional neural network (CNN). Here we have used data augmentation with CNN model and also, we have made a comparative study with Multilayer Perceptron (MLP) and Single Layer Perceptron (SLP). The model is tested using 4480 images with 20 subjects where we have found the accuracy of CNN is 82.75% and CNN with data augmentation is 98.33%, SLP is 39.91% and MLP is 63.57%. Based on the comparative study CNN with data augmentation proves the best performance.