Semester

Fall

Date of Graduation

2012

Document Type

Thesis

Degree Type

MS

College

Statler College of Engineering and Mineral Resources

Department

Lane Department of Computer Science and Electrical Engineering

Committee Chair

Xin Li

Committee Co-Chair

Hanny Ammar

Committee Member

Guodong Guo

Abstract

Recognizing Arabic handwritten words is a difficult problem due to the deformations of different writing styles. Moreover, the cursive nature of the Arabic writing makes correct segmentation of characters an almost impossible task. While there are many sub systems in an Arabic words recognition system, in this work we develop a sub system to recognize Part of Arabic Words (PAW). We try to solve this problem using three different approaches, implicit segmentation and two variants of holistic approach. While Rothacker found similar conclusions while this work is being prepared, we report the difficulty in locating characters in PAW using Scale Invariant Feature Transforms under the first approach. In the second and third approaches, we use holistic approach to recognize PAW using Support Vector Machine (SVM) and Active Shape Models (ASM). While there are few works that use SVM to recognize PAW, they use a small dataset; we use a large dataset and a different set of features. We also explain the errors SVM and ASM make and propose some remedies to these errors as future work.

Share

COinS