We found a match
Your institution may have access to this item. Find your institution then sign in to continue.
- Title
EFFECT OF THE TRESHOLDING TO SIGNATURE VERIFICATION USING ARTIFICIAL NEURAL NETWORK.
- Authors
Çıkoğlu, Serdar; Temurtaş, Fevzullah; Yumuşak, Nejat
- Abstract
In this study, signature verification has been done using artificial neural networks. In signature verification process firstly, signatures have been normalized and thresholded in order to prevent the noises in background. Then the features used in artificial neural network's learning have been extracted. Signatures have been differentiated using five features such as signature density, horizontal relative difference between signature centers, vertical relative difference between signature centers, signature width, signature height. 15 features have been formed from these features using three different threshold calculations. So three types of articial neural network model using this 5,10 and 15 feature for signature verification have been used. Effects of this thresholds were investigated on learning performance of the neural network by using 3 different tresholding (185,170,150).
- Subjects
VERIFICATION of computer systems; THRESHOLD (Perception); SIGNATURES (Writing); ARTIFICIAL neural networks; LEARNING; NOISE
- Publication
Teknoloji, 2004, Vol 7, Issue 1, p151
- ISSN
1302-0056
- Publication type
Article