A New Feature Extraction Method for TMNN-Based Arabic Character Classification

Authors

  • Khalid Saeed
  • Majida AlBakoor

Keywords:

Arabic characters, Backpropagation neural networks, Toeplitz matrices

Abstract

This paper describes a hybrid method of typewritten Arabic character recognition by Toeplitz Matrices and Neural Networks (TMNN) applying a new technique for feature selecting and data mining. The suggested algorithm reduces the NN input data to only the most significant and essential-for-classification points. Four items are determined to resemble the distribution percentage of the essential feature points in each part of the extracted character image. Feature points are detected depending on a designed algorithm for this aim. This algorithm is of high performance and is intelligent enough to define the most significant points which satisfy the sufficient conditions to recognize almost all written fonts of Arabic characters. The number of essential feature points is reduced by at least 88 %. Calculations and data size are then consequently decreased in a high percentage. The authors achieved a recognition rate of 97.61 %. The obtained results have proved high accuracy, high speed and powerful classification.

Downloads

Download data is not yet available.

Downloads

Published

2012-01-27

How to Cite

Saeed, K., & AlBakoor, M. (2012). A New Feature Extraction Method for TMNN-Based Arabic Character Classification. Computing and Informatics, 26(4), 403–420. Retrieved from http://147.213.75.17/ojs/index.php/cai/article/view/317