On Feature Extraction and Selection for Arabic Character Recognition PDF

Keywords

feature extraction, Arabic characters, engineering

Abstract

An algorithm to select an optimum set of density features for the Arabic character set is developed and implemented. The method involves generation of all possible features by means of position-by-position matching between characters. The feature selected, each have a correlation coefficient greater than a certain threshold. The obtained optimum feature set is utilized in the classification stage employing a sequential tree search technique. Computational results are presented and discussed.