Recognize printed Arabic letter using new geometrical features

Abstract

The task of recognizing the shape of Arabic letters using modified algorithms discussed in this paper. The difficulty of recognizing these letters is summarized in the shape of the Arabic letter within a word from a large set of letters has a similar shape. Moreover, the shape of the letter is different depending on its position begin, middle, end within a word. Therefore, it is necessary to introduce new geometric features to categorize each letter. The suggested algorithm with 19 features is used in this paper. These features, like define points for each letter, divide a letter to blocks, edge detection and other features are shown in the suggested algorithm. The introduced geometric features give a high accuracy to recognize printed Arabic letter within a word or text. Minimum distance criteria used to estimate the error of the recognition process between the database and the tested Arabic letter. This method is good to explain the behaviour of the designed algorithm code to distinguish the geometric properties and the accuracy reaches 99.8% for the proposed method. The letter size changes geometry details when the font size is changed. The studied font is Times New Roman with size 30, 36, and 39.