Identifikasi Tanda Tangan Berdasarkan Grid Entropy Menggunakan Multi Layer Perceptron

Abstract

The signature is one frequently proof validation used on documents. Recognition of signature is required to verify document whether the signature is gived by concerned person or others. In this study the authors design a signature identification system based on the value of entropy that taken from the grid image of an image of a signature. Training and testing model using a multi layer perceptron and cross validation by three grid sizes (4x4, 8x8, and 16x16) and two types of image representation (binary image and the image of the outline). The best test results obtained on the grid size 8x8 using outline image that is the accuracy rate of 97.78%, the value of the correlation 0.981, and a kappa value of 0.977.