Bio-inspired multilayered and multilanguage Arabic script character recognition system

Handwritten character recognition is one of the most fascinating, complex and challenging task of pattern recognition research and it has been an ongoing research since the early days of computer. The aim of character recognition is to imitate the human reading capabilities to machine so that it can act like human. Although it is a very difficult task, it has several practical application. Naturally translation, rotation and position are independent of human visual perception which cannot be easily attained by computer system, but these have also some effect on human reading speed and accuracy. Bio-inspired pattern recognition has proved itself a powerful method to deal with complex patterns whereas fuzzy logic is powerful classifier to recognize irregular and implement human reasoning concept. We combine the human visual concept with fuzzy logics to involve the human reasoning for multilanguage Arabic script based language character recognition. The proposed approach is tested on Multilanguage only written in Nasta'liq style and it provides 86.2% accuracy. © 2012 ICIC International.