Arabic Hand Written Character Recognition Using Modified Multi-Neural Network

Hand written recognition is an interesting area of current artificial intelligence and advanced computing’s researchers. The complexity of the language controls the ability and the challenge of recognition its characters, whereas this complexity and uncertainty becomes multiplied. The use of Latin languages like English, or Spanish, limits the uncertainty because of the limited structure of the character. Arabic language characters are very complex in comparison with the most languages in the world. The character in the Arabic language is not static, it has many shapes – in almost – depends on it location in the world, in addition to the convention. More complexity of that, the Arabic characters is being writing continuously connecting the character with the next or the previous – in most –. This research adopted an algorithm to recognize the single segment characters, supposing that, the connected characters are separated and already segmented. This faces all problems of the character shape, location, style, and the user’s style of writing. This paper implements a neural network based system that uses cascaded networks to recognize the characters. MATLAB program is designed to test the implementation and recording the results.