Kurdish Handwritten character recognition using deep learning techniques

Rebin M Ahmed; Tarik A Rashid; Polla Fattah; Abeer Alsadoon; Nebojsa Bacanin; Seyedali Mirjalili; S Vimal; Amit Chhabra

doi:10.1016/j.gep.2022.119278

Kurdish Handwritten character recognition using deep learning techniques

Gene Expr Patterns. 2022 Dec:46:119278. doi: 10.1016/j.gep.2022.119278. Epub 2022 Oct 3.

Authors

Rebin M Ahmed¹, Tarik A Rashid², Polla Fattah³, Abeer Alsadoon⁴, Nebojsa Bacanin⁵, Seyedali Mirjalili⁶, S Vimal⁷, Amit Chhabra⁸

Affiliations

¹ IT Department, Faculty of Aplied Science, Tishk International University, Erbil, Iraq. Electronic address: rebin.mohammed@tiu.edu.iq.
² Computer Science and Engineering Department, University of Kurdistan Hewler, Erbil, Iraq. Electronic address: tarik.ahmed@ukh.edu.krd.
³ Software and Informatics Engineering, Salahaddin University-Erbil, Erbil, Iraq. Electronic address: pollaeng@gmail.com.
⁴ School of Computing Engineering and Mathematics, Western Sydney University, Sydney City Campus, Australia; Asia Pacific International College (APIC), Information Technology Department, Sydney, Australia; Kent Institute Australia, Information Technology Department, Sydney, Australia. Electronic address: alsadoon.abeer@gmail.com.
⁵ Singidunum University, Belgrade, Serbia. Electronic address: nbacanin@singidunum.ac.rs.
⁶ Centre for Artificial Intelligence Research and Optimisation, Torrens University, Australia; Yonsei Frontier Lab, Yonsei University, Seoul, South Korea. Electronic address: ali.mirjalili@gmail.com.
⁷ Department of Artificial Intelligence and Data Science, Ramco Institute of Technology, North Venganallur Village, Rajapalayam 626 117, Virudhunagar District, Tamilnadu, India. Electronic address: svimalphd@gmail.com.
⁸ Department of Computer Engineering and Technology, Guru Nanak Dev University, Amritsar India. Electronic address: amit.cse@gndu.ac.in.

PMID: 36195308
DOI: 10.1016/j.gep.2022.119278

Abstract

Handwriting recognition is regarded as a dynamic and inspiring topic in the exploration of pattern recognition and image processing. It has many applications including a blind reading aid, computerized reading, and processing for paper documents, making any handwritten document searchable and converting it into structural text form. High accuracy rates have been achieved by this technology when recognizing handwriting recognition systems for English, Chinese Arabic, Persian, and many other languages. However, there is not such a system for recognizing Kurdish handwriting. In this paper, an attempt is made to design and develop a model that can recognize handwritten characters for Kurdish alphabets using deep learning techniques. Kurdish (Sorani) contains 34 characters and mainly employs an Arabic/Persian based script with modified alphabets. In this work, a Deep Convolutional Neural Network model is employed that has shown exemplary performance in handwriting recognition systems. Then, a comprehensive database has been created for handwritten Kurdish characters which contain more than 40 thousand images. The created database has been used for training the Deep Convolutional Neural Network model for classification and recognition tasks. In the proposed system the experimental results show an acceptable recognition level. The testing results reported an 83% accuracy rate, and training accuracy reported a 96% accuracy rate. From the experimental results, it is clear that the proposed deep learning model is performing well and comparable to the similar to other languages handwriting recognition systems.

Keywords: Convolutional neural network; Database; Kurdish alphabet; Offline handwriting recognition.

MeSH terms

Algorithms
Deep Learning*
Handwriting
Neural Networks, Computer
Pattern Recognition, Automated* / methods