Title: Cursive multilingual characters recognition based on hard geometric features
Authors: Amjad Rehman; Majid Harouni; Tanzila Saba
Addresses: Artificial Intelligence and Data Analytics (AIDA), Lab CCIS, Prince Sultan University, Riyadh, Saudi Arabia ' Department of Computer Science, Dolatabad Branch Islamic Azad University, Isfahan, Iran ' Artificial Intelligence and Data Analytics (AIDA), Lab CCIS, Prince Sultan University, Riyadh, Saudi Arabia
Abstract: The cursive nature of Arabic script characters segmentation and recognition has attracted researchers from academia and industry. However, despite several decades of research, still Arabic characters classification accuracy is not up to the mark. This paper presents an automated approach for Arabic characters segmentation and recognition. The proposed methodology explores character's boundaries based on their geometric features, prior to their recognition. However, due to uncertainty and without dictionary support few characters are over-divided. To expand the productivity of the proposed methodology a hybrid neural model (HNN) is proposed. The model is composed of trained MLP and RBF networks to assist the character's boundary recognition process. For reasonable examination, only benchmark dataset is utilised.
Keywords: Arabic character recognition; features mining; geometrical features; hybrid neural model; HNN.
DOI: 10.1504/IJCVR.2020.107244
International Journal of Computational Vision and Robotics, 2020 Vol.10 No.3, pp.213 - 222
Received: 26 Sep 2018
Accepted: 22 Apr 2019
Published online: 11 May 2020 *