Postal Code Handwritten Recognition System
Loading...
Date
2016-06-10
Journal Title
Journal ISSN
Volume Title
Publisher
University of M'sila
Abstract
A three based classifiers system was created. A back-propagation neural network With one hidden layer , a support Vector machine classifier, and K- nearest Neighbor classifier were used to create an adaptive postal code digits recognition system by using the Hu moments invariants feature extraction method. The system was trained and evaluated through different forms of handwriting samples provided by both male and female participants. Experiments tested, the effect of the size set on the recognition accuracy, and the effect of handwriting style on the recognition accuracy. Results showed that the handwriting style of the subjects had varying and drastic effects on the recognition accuracy which allowed to identify some of the problems With the system digits encoding.
Description
Keywords
KNN, SVM, ANN, Hu moments ,code postal, recognition system.