Character Recognition using Backpropagation Neural Network for Printed Documents
Digital documents from the scanner device cannot be edited. To be able to edit digital documents, Optical Character Recognition (OCR) technology is needed. This research was conducted with the aim of implementing backpropagation artificial neural networks in printed documents and to find out how the accuracy of the implementation of backpropagation artificial neural networks in printed documents. This research uses multilayer networks with three layers. The input layer consists of 225 nodes with 15 × 15 pixels digital image as input, hidden layer consists of 110 nodes, and the output layer consists of 54 nodes representing A-Z, a-z, point punctuation (.), and comma punctuation (,). The learning rate used in this research is 0,29. The average accuracy level obtained from the implementation of backpropagation artificial neural networks in this research was 94 % for Ms Arial Unicode font type, 96,6 % for Tahoma font type, and 94 % for Times New Roman font type.
This work is licensed under a Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License.