Jurnal Publikasi STMIK Pontianak

Penerapan Algoritma Jaringan Saraf Tiruan Backpropagation untuk Pengenalan Pola Tanda Tangan

Abstrak

This paper discusses the process of signature identification using computer by implementing the artificial neural network of backpropagation and the function of activation using sigmoid biner by initializing small random numbers for weights and biases, and value range generated between 0 and 1. The method of analysis and its design use Object-Oriented Analysis and Design (00AD) through the approach of Prototype and its test uses black box. This application receives the input of signature pattern using digital image in the format of bmp, jpg, png, gif and tif sized in 300 x 300 pixels or the signature pattern directly drawn by using mouse in a specified area. The architecture of artificial neural network of backpropagation is 400 nodes of input layer, 50 nodes of hidden layer and 4 nodes of output layer.

Keywords: Signature pattern, artificial neural network of backpropagation, sigmoid biner. object-oriented analysis and design, prototype, black box.