| dc.description.abstract | Signature identification is a process for identifying a person who has the signature. Nowadays, there are many signature forgeries. A system that can recognize signature patterns is required to overcome these problems. Identification system which is implemented in this research uses Probabilistic Neural Networks (PNN). It works faster than other neural network approaches such as backpropagation because PNN approach only needs one iteration in training process. This research uses 200 images of signatures which consist of 20 signatures of each 10 respondents. This research is divided into two parts which are images without Wavelet decomposition and iamges with Wavelet decomposition. The results of this research are the system using Wavelet decomposition level 2 and level 3 in 4-fold cross validation has 98% accuracy. Different training data causes different test results. The greater the number of training data tends to increase the accuracy. Computation time is faster with the Wavelet decomposition and even faster with the addition of decomposition level. Keyword : Probabilistic Neural Networks (PNN), signature, Wavelet. | id |
