Title: Pixel distribution-based features for offline Arabic handwritten word recognition

Authors: Qais Ali Al-Nuzaili; Siti Z. Mohd Hashim; Faisal Saeed; Mohammed Sayim Khalil; Dzulkifli Bin Mohamad

Addresses: Faculty of Computing, Universiti Teknologi Malaysia, Johor, Skudai – 81310, Malaysia ' Faculty of Computing, Universiti Teknologi Malaysia, Johor, Skudai – 81310, Malaysia ' Faculty of Computing, Universiti Teknologi Malaysia, Johor, Skudai – 81310, Malaysia ' Center of Excellence in Information Assurance, CoEIA, King Saud University, Riyadh, Saudi Arabia ' Faculty of Computing, Universiti Teknologi Malaysia, Johor, Skudai – 81310, Malaysia

Abstract: Handwritten word recognition is the ability of a computer to receive and interpret intelligible handwritten input. An important document recognition application is bank cheque processing. The Arabic bank cheque processing system has not been studied as much as Latin and Chinese systems. The domain of handwriting in the Arabic script presents unique technical challenges; proposing a model for feature extraction which combines multiple types of features most likely will help to improve the recognition rate. This work proposed a pixel distribution-based features model (PDM) for offline Arabic handwritten word recognition. Two combination levels were used: the first combines different features and the second combination was done by ensemble classifiers. The AHDB dataset was used, and the experimental results showed superior performance when combining multiple features and using multi classifiers.

Keywords: feature extraction; Arabic handwriting recognition; Arabic legal amounts; majority voting; holistic approach; pixel distribution-based features; word recognition; bank cheque processing; Arabic bank cheques; ensemble classifiers.

DOI: 10.1504/IJCVR.2017.081243

International Journal of Computational Vision and Robotics, 2017 Vol.7 No.1/2, pp.99 - 122

Received: 22 Dec 2014
Accepted: 08 Apr 2015

Published online: 07 Dec 2016 *

Full-text access for editors Access for subscribers Purchase this article Comment on this article