\ Offline Handwritten Digit Recognition Using Neural Network

Offline Handwritten Digit Recognition Using Neural Network

Rıza Tuğrul Kale

Tarih ve Saat

15 Eylül 2015 - 16:00


Mühendislik Fakültesi - A511

This study presents an Artificial Neural Network implementation for offline classification of handwritten digits which are obtained from popular NIST (National Institute of Standards and Technology) database.
Character or digit recognition can be effortlessly done by humans, whether the character is of any magnitude and form. But it is a challenging task in the computer vision. There are numerous studies for digit recognition and competitions for the classification of digits with minimum error rate.
Feature selection of raw data (training set) is the most important factor to simplify the input of the network. Besides, this provides an easy way to handle the big data set.
In this study, feature selection methods such as image reduction, Principal Component Analysis and Artificial Neural Network as a classification technique have been implemented using MATLAB.
The study has been conducted on the training set which contains 42,000 handwritten digits with 28 by 28 pixel size. After feature selection and data reduction, 28,000 new test digits  have been predicted by using Neural Network classifier.

Yeditepe Üniversitesi, Endüstri ve Sistem Mühendisliği
26 Ağustos Yerleşimi, Kayışdağı Cad. 34755 Ataşehir, İstanbul

(216) 578 04 50 info@sye.yeditepe.edu.tr