IDENTIFIKASI TANDA-TANGAN MENGGUNAKAN JARINGAN SARAF TIRUAN PERAMBATAN-BALIK (BACKPROPAGATION)
Abstract: Human signature
identification is a process for identifying and obtaining a person who has the
signature. Signature identification technology includes in biometrics system
which uses a behavioral human nature characteristics. For the time being, there
are many signature forgeries which are generally make a harm for people who
have the signatures. Signature forgery occurs easily for which a system which
can assist to identify a person’s signature is required. Identification system
which will be implemented uses Backpropagation Neural Network model and is
supprorted by Delphi programming language. In order to identify a signature,
image of signature firstly needs a preprocessing and features extraction. In
the preprocessing, there are three stages which have to be performed, there
are: converting the grayscaled image, contrasting the image, and edge detection
of the image. Features extraction process is performed by segmenting the image
in the form of rows and columns which has a purpose to get a significant
feature information of the image of signature, as well as to get a data value
which will be an input for neural network. Neural network training is performed
to get an accurate classification from trained data input of signatures. A
signature can be identified when the signature is comprised in one of classes
which formed of training process. The research uses 150 images of signatures
which consist of 10 responders for database for which it requires 10 data from
each responder and 5 responders from outer side of database for which it
requires 5 data from each responder. Conclusions of the research are that the
application system has a 95% percentage of success level for identifying the
signatures from the testing of trained data, while it has only 88% percentage
of success level from the testing of outer side of database.
Keywords: Signature
identification, biometrics, feature extraction, backpropagation neural network
Penulis: Hidayatno, A. and Isnanto,
R.R. and Buana, D.K.W.
Kode Jurnal: jptinformatikadd080004
