PENGENALAN KARAKTER PADA SURAT MASUK MENGGUNAKAN NEURAL NETWORK BACKPROPAGATION
Abstract: A letter
is a means
of communication to convey
information in writing by one party to the
other party. It
functions include five
things: means notices, requests,
thoughts and ideas.
A formal letter is a
letter that contain
the official or a particular
business problem. Along with the increasing
demand for the
current document digitalization, OCR
(Optical Character Recognition) application
is often used
to identify the image of characters
then will be converted into text files. Furthermore,
Artificial Neural Network is
one kind of
technology approach that promising
can increase the computer's ability
to recognize and
represent pattern.
Meanwhile, Backpropagation is one
kind of
neural network algorithm
which common used. Therefore,
a character recognition application
using neural network backpropagation will
be designed. Application need a
data image of
document letter as an input, it’s
obtained from scanning,
then continued by pre-processing, segmentation, normalization size process.
After all, it’s ready to be processed
in network and
will get characters text
as an output.
Result of the
test had been showed
that, the best
recognition accuracy rate on
the characters trained,
font Arial, with learning rate (α) 0.2,
momentum (µ) 0.5 and 15
epoch is 71.95%
and for the characters without
trained, font Times
New Roman and Courier
New, the recognition accuracy rate
is 40.24%. Overall,
the result of percentage
success the recognition
of 790 characters in 15 document
letter is 84.56%.
Keywords: character recognition, neural network, backpropagation, pattern recognition, optical
character recognition
Penulis: Rizqia Lestika Atimi
Kode Jurnal: jptinformatikadd130170
