Date of Award

2010

Publication Type

Master Thesis

Degree Name

M.A.Sc.

Department

Electrical and Computer Engineering

First Advisor

Maher Sid-Ahmed

Keywords

Applied sciences

Rights

info:eu-repo/semantics/openAccess

Creative Commons License

Creative Commons Attribution-Noncommercial-No Derivative Works 4.0 License
This work is licensed under a Creative Commons Attribution-Noncommercial-No Derivative Works 4.0 License.

Abstract

The subject of handwritten character recognition has been receiving considerable attention in recent years due to the increased dependence on computers. Several methods for recognizing Latin, Chinese as well as Kanji characters have been proposed. However, work on recognition of Arabic characters has been relatively sparse. Techniques developed for recognizing characters in other languages can not be used for Arabic since the nature of Arabic characters is different. The shape of a character is a function of its location within a word where each character can have two to four different forms. Most of the techniques proposed to date for recognizing Arabic characters have relied on structural and topographic approaches.

This thesis introduces a decision-theoretic approach to solve the problem. The proposed method involves, as a first step, digitization of the segmented character. The secondary part of the character (dots and zigzags) are then isolated and identified separately thereby reducing the recognition issue to a 20 class problem or less for each of the character forms. The moments of the horizontal and vertical projections of the remaining primary characters are calculated and normalized with respect to the zero order moment. Simple measures of shape are obtained from the normalized moments and incorporated into a feature vector. Classification is accomplished using quadratic discriminant functions. The approach was evaluated using isolated, handwritten characters from a data base established for this purpose. The classification rates varied from 97.5% to 100% depending on the form of the characters. These results indicate that the technique offers significantly better classification rates in comparison with existing methods.

Share

COinS