Counter - propagation network for character recognition

Trần Ngọc Hà, Nguyễn Thanh Thủy

Abstract


Much effort has been extended in making computer recognize characters automatically. Methods currently widely used for character recognition are mainly involved in pattern marching using image processing techniques. One limitation of such methods is their inability to respond to variations.

This paper presents the use of counter - propagation network based on Kohonen network for character recognition. A first attempt at recognizing 37 symbols has been carried out. Obtained results showed that the system is very robust, stable and still capable to retrieve correctly characters for 25% corrupted images.




DOI: https://doi.org/10.15625/1813-9663/14/2/7894

Journal of Computer Science and Cybernetics ISSN: 1813-9663

Published by Vietnam Academy of Science and Technology