Effect of system parameters on feature extraction sets for Arabic handwritten text recognition

Abstract

The purpose of this study is to analyze the effect of different parameters on a Pattern Recognition System for Arabic Handwritten Text Recognition, and perform different experimental tests in order to obtain the optimum values for the process. In the first introductory section, source data material and the tools used in this work are introduced and explained. The thesis then focuses on the Feature Extraction Process, providing details about different strategies or methods that can be used on the process. In the experimental section, the most important test results are given and the variable parameters are individually analyzed. Finally, different combination schemes are implemented in order to prove the effectiveness of the Slanted Windows. The results provide some support for the correct selection of parameter values for future implementations of the system. However, the optimum parameter values should not be considered as absolute values, due to the fact that the aim is to guide the researchers for future implementations of the system.