Presenting a decision-tree-based method for recognition of persian handwritten numbers

This paper presents an effective method for recognition of Persian handwritten numbers. The method is based on structural features and decision-tree learning technique. The presented method includes three stages: First a preprocessing is made on the number, in this way the entry number has been thinned down while at the same time some noise coming as a result of lack of congruence in numbers thickness is deleted. In the second stage four properties applied at the stage of recognition have been extracted. The third stage comprises of creating a decision genealogy (tree) related to the order of classification. The total dataset used in the experimental stage is the database of Hoda documented upon by researchers. The experiment's results indicate the efficacy on the part of the method suggested for recognizing handwritten Persian numbers.