Adi Slamet Kusuma Wardana (2011) Pengenalan Pola Huruf Tulisan Tangan Menggunakan Jaringan Syaraf Backpropagation. Skripsi thesis, UNIVERSITAS AIRLANGGA.

	Text (HALAMAN JUDUL) 1. HALAMAN JUDUL.pdf Download (127kB)
	Text (ABSTRAK) 2. ABSTRAK .pdf Download (88kB)
	Text (DAFTAR ISI) 3. DAFTAR ISI .pdf Download (99kB)
	Text (BAB 1 PENDAHULUAN) 4. BAB I PENDAHULUAN .pdf Download (128kB)
	Text (BAB II TINJAUAN PUSTAKA) 5. BAB II TINJAUAN PUSTAKA .pdf Restricted to Repository staff only until 19 April 2023. Download (477kB) | Request a copy
	Text (BAB III METODE PENELITIAN) 6. BAB III METODE PENELITIAN.pdf Restricted to Repository staff only until 19 April 2023. Download (198kB) | Request a copy
	Text (BAB IV PEMBAHASAN) 7. BAB IV PEMBAHASAN .pdf Restricted to Repository staff only until 19 April 2023. Download (535kB) | Request a copy
	Text (BAB V PENUTUP) 8. BAB V PENUTUP.pdf Restricted to Repository staff only until 19 April 2023. Download (42kB) | Request a copy
	Text (DAFTAR PUSTAKA) 9. DAFTAR PUSTAKA .pdf Download (45kB)
	Text (LAMPIRAN) 10. LAMPIRAN .pdf Restricted to Repository staff only until 19 April 2023. Download (449kB) | Request a copy

Abstract

Pattern recognition of handwritten letters is a topic that has been researched for many ears. The problems encountered in pattern recognition of handwritten characters is very complex, including the variety of models of handwriting and handwriting size. One of Pattern recognition of handwritten letters is artificial neural network, where this method uses a similar principle workings of the human brain. The purpose of this final project is applying artificial neural network to Pattern recognition of handwritten letters and create a program simulating this method using Visual Basic 6.0 software with supporting operation system. Artificial neural network architecture used is multilayer neural network with backpropagation algorithm. Data used are handwritten letters images with 60 x 60 pixel size which transformed into numeric with image processing. From the image processing numerical values obtained in the form of initial matrix size 60 x 60, with the segmentation process change initial matrix into a matrix measuring 20 x 20, then with the normalization matrix is converted into the final matrix size of 400 x 1 for each picture. From the normalization process will be a backpropagation neural network input for pattern recognition of handwritten letters. After the normalization process, input will be processed for training and testing. Network training using 156 handwritten letters data with 0,9 learning rate and 0,001 error, looping stopped at 143685th iteration. Validation test results for 104 images, we concluded that 71.15% of all successful validation images well recognized

Actions (login required)

View Item