Design and implementation of a system for recognizing Arabic handwritten words with learning ability
Kamal Mansour Jambi · 1992
Arabic character recognition is an open field. There are some properties for Arabic characters that make their recognition more challenging. They are written cursively with often overlapping. Some groups are having the same main body with a slight change in the number of dots and their position. The shape of a character changes depending on its location within the word. This implemented system recognizes Arabic handwritten words. It goes through some preprocessing operations such as thinning, finding the extreme boundaries, and resolving the overlapping. Segmentation of the word into its characters is another problem to deal with. The approach adopted here tries to associate between the feature points and one of the six windows that are generated by the frame enclosing each character under consideration. The obtained results and some interpretations of them are presented. A discussion on how to improve these results is given. The characteristics of this system such as simplicity, ability to learn and interactive segmentation are elaborated. Some considerations for future research are also provided.