Deep convolutional neural network for isolated Arabic handwritten character recognition: design, evaluation, and comparative study

Eyad Talal Attar ¹

⁰Department of Electrical and Computer Engineering, Faculty of Engineering, King Abdulaziz University, Jeddah, 21589, Saudi Arabia. etattar@kau.edu.sa.

Scientific Reports +

|

November 27, 2025

View abstract on PubMed

Summary

A new deep Convolutional Neural Network (CNN) accurately recognizes handwritten Arabic letters, achieving 96.8% accuracy. This advanced model significantly outperforms traditional methods for Arabic handwriting recognition tasks.

Area Of Science

Computer Science
Artificial Intelligence
Machine Learning

Background

Handwritten Arabic character recognition is challenging due to script complexity.
Traditional methods struggle with cursive forms, positional variations, and diverse styles.

Purpose Of The Study

To develop a deep Convolutional Neural Network (CNN) for classifying isolated handwritten Arabic letters.
To evaluate the CNN's performance against classical machine learning techniques.

Main Methods

A tailored deep Convolutional Neural Network (CNN) architecture was proposed.
A balanced dataset of 28 Arabic alphabet classes was used with preprocessing and augmentation.
Performance was assessed using cross-validation, confusion matrix analysis, and statistical testing.

Main Results

The proposed CNN achieved a classification accuracy of 96.8%.
This significantly outperformed Support Vector Machine (SVM) at 85.3% and K-Nearest Neighbors (KNN) at 82.1%.
A paired t-test confirmed the statistical significance of the CNN's superiority (p < 0.01).

Conclusions

The developed CNN offers a highly accurate solution for isolated handwritten Arabic character recognition.
The model shows potential for applications like document digitization and assistive technologies.
The architecture is extensible to other cursive-based languages and future work includes connected script recognition.

Keywords:

Convolutional neural network (CNN) Cursive script Deep learning Handwritten arabic character recognition Image classification Multilingual recognition Optical character recognition (OCR) Pattern recognition Statistical validation