IDENTIFIKASI DAN VERIFIKASI TANDA TANGAN STATIK MENGGUNAKAN BACKPROPAGATION DAN ALIHRAGAM WAVELET

KUMALASANTI, ROSALIA ARUM (2015) IDENTIFIKASI DAN VERIFIKASI TANDA TANGAN STATIK MENGGUNAKAN BACKPROPAGATION DAN ALIHRAGAM WAVELET. S2 thesis, UAJY.

	Text (Halaman Judul) MTF002014.pdf Download (222kB)
	Text (Bab I) MTF102014.pdf Download (44kB)
	Text (Bab II) MTF202014.pdf Download (123kB)
	Text (Bab III) MTF302014.pdf Restricted to Registered users only Download (64kB)
	Text (Bab IV) MTF402014.pdf Restricted to Registered users only Download (792kB)
	Text (Bab V) MTF502014.pdf Download (50kB)

Abstract

The signature is an important biometric attributes of a person or individual who can be used as identification. Signature is a common and traditional ways are often used as a valid ID. This makes the presence of a signature to be important, so we need a system that used to provide security in order not to be abused by irresponsible parties. Various approaches have been proposed in the development of identification and verification of signatures which aims to minimize fraud that forged the signature. This study will be discussed on the identification and verification of signatures for authenticity. This process consists of two main parts, training and testing phase. Image size is 256x256 pixel used. In training phase, the image of the signature subject to several prosesses that threshold, the transformation wavelet, normalized and then be trained by using algorithm Artificial Neural Network (ANN) backpropagation. The testing phase has the same process as in the training phase but at the end of the process will be a comparison the image data that has been stored with the image comparison. ANN can perform optimally when trained using input data that has been taken into consideration the size, parameters, and the numbers of nodes on the network. optimal results are obtained by using a neural network has two hidden layers, each of 20 and 10 nodes, the transformation of Haar wavelet at level 4 and learning rate of 0,12. Training and testing in the identification phase, each providing an accuracy of 95,56% and 100%. Training and testing on the verification phase, each providing an accuracy of 100% and 96,67%

Actions (login required)

View Item