Pengenalan Dokumen Jawa Menggunakan Jaringan Syaraf Tiruan Feedforward Elman Type Algorithm

Abstract

The training process is the most important thing to recognize something, because without the process training, it’s impossible to recognize something with what datas you have. The training process of Javanese letter has its own difficulty level for collecting the extraction feature datas that can be used for training process and it took a long time to do the training process. In this paper developed an application that can make the process training to recognize of Javanese Letter with the extration feature datas.

The process is carried out as follows: image that has been previously segmented, then extract the feature datas and used it as input to begin the training process with  Elman Type. The Inputs are 88 atribute from data extraction features, that is the result of extraction features based on area-based theory from a picture of Javanese letter.  The result of the training process are some datas,  bias, weight datas, when the data is deemed to able produce good output, then it can be used for recognize the Javanese letter from the other source extraction feature datas . The output is some datas, like bias, weight, and Java letter recognition results stored in file.txt. This application is made to the programming language C # with Microsoft Visual Studio 2010 as the IDE.

The results show that javanese letter recognition accuracy by using Elman reach 85.16 % .