Widodo, IsnawanTri (2011) Pengenalan Kata Tulisan Tangan Menggunakan Jaringan Syaraf Tiruan Optical Backpropagation. Sarjana thesis, Universitas Brawijaya.
Abstract
Sistem pengenalan tulisan tangan telah banyak diimplementasikan untuk aplikasi pengenalan tulisan tangan, antara lain sistem pemilahan surat otomatis, pengolahan cek di bank dan banyak aplikasi lainnya. Aplikasi pengenalan tulisan tangan banyak menggunakan metode jaringan syaraf tiruan. Hal ini karena tingkat akurasi yang dimiliki cukup tinggi. Optical Backpropagation adalah salah satu algoritma dalam jaringan syaraf tiruan yang didesain untuk menyelesaikan beberapa kendala yang berkaitan dengan backpropagation standar. Salah satu aspek penting dari algoritma ini adalah kemampuannya untuk menghindari local minima dalam proses pelatihannya dengan kecepatan yang cukup tinggi untuk mencapai konvergensi selama proses pelatihan. Dalam penelitian ini, sampel kata yang didapat akan disegmentasi menjadi karakter mengunakan algoritma Modified Vertical Histogram. Dari karakter yang tersegmentasi tersebut, kemudian dilakukan proses pelatihan dan pengenalan. Dari penelitian ini diperoleh struktur jaringan syaraf tiruan Optical Backpropagation dengan jumlah neuron pada input layer sebanyak 400 unit, jumlah neuron pada hidden layer sebanyak 100 unit, jumlah neuron pada output layer sebanyak 400 unit, nilai learning rate sebesar 0.004 dengan max epoch sebesar 5000. Jaringan syaraf tiruan yang terbentuk mampu mengenali citra tulisan tangan sebesar 96,70% untuk citra yang berisi huruf tunggal yang pernah dilakukan pembelajaran dan hasil pengenalan terhadap citra yang berisi huruf tunggal yang belum dilakukan pembelajaran sebesar 56.41%. Sedangkan tingkat keakuratan sistem secara keseluruhan terhadap kata yang diujikan sebesar 72.45%.
English Abstract
Handwriting recognition system has been implemented for handwriting applications such as automatic mail voting system, signature validation, processing check at the bank and many others. Some of handwriting application systems used artificial neural network method. That was because of its high level of accuracy. Optical Back propagation is one of the algorithms in artificial neural networks designed to overcome several problems associated with standard back propagation. One of the important aspects of this algorithm was its ability to avoid local minima in the training process with high speed level to achieve convergence during the training process. In this research, handwriting samples obtained will be segmented into characters using a Modified Vertical Histogram algorithm. From this segmented character, the training and recognizing process will be done. From this study was obtained the structure of Optical Back propagation neural network with the number of neurons in the input layer were 400 units, the number of neurons in the hidden layer were 100 units, the number of neurons in the output layer were 400 units, the value of learning rate were 0004 with the max epoch were 5000. Artificial neural network that has formed was capable to recognize handwriting image were 96.70% for images containing a single letter has ever done learning and recognition results to images that contain single letter that has ever done learning were 56.41%. While, the overall level of accuracy of the system is tested with the word were 72.45%.
| Item Type: | Thesis (Sarjana) |
|---|---|
| Identification Number: | SKR/MIPA/2011/214/051103321 |
| Subjects: | 500 Natural sciences and mathematics > 510 Mathematics |
| Divisions: | Fakultas Matematika dan Ilmu Pengetahuan Alam > Matematika |
| Depositing User: | Unnamed user with email repository.ub@ub.ac.id |
| Date Deposited: | 28 Oct 2011 09:14 |
| Last Modified: | 22 Oct 2021 07:50 |
| URI: | http://repository.ub.ac.id/id/eprint/152667 |
Preview |
Text
051103321.pdf Download (4MB) | Preview |
Actions (login required)
 |
View Item |