APPLICATION OF ARTIFICIAL NEURAL NETWORKS IN PATTERN RECOGNITION

Abstract

This thesis introduces new techniques for Arabic handwritten character recognition systems. First of them is C.C.D. (Connected Components Detector), which is applied on Japanese handwritten characters [5], and is applied for the first time on Arabic handwritten characters in this thesis. C.C.D. is based on transforming the original . character image to 4 sub-images. These 4 sub-images are the results of vertical, horizontal, +45, and -45 scans of the original image. C.C.D. Feature extraction procedure is applied on each sub-image, which is counting the number of connected (successive) components with the same type i.e. number of l's and number of O's in each raw then save that results in a matrices which represents the features extracted.