Advancements in CNN Architectures for Offline Handwritten Arabic Character Recognition

Abstract

Analyzing and classifying images of Arabic handwritten characters is crucial for text understanding and interpretation from image data. The recognition of handwritten Arabic characters not only preserves the integrity of the Arabic language but also enhances computer vision applications tailored for Arabic script. Existing literature often proposes complex architectures, which can hinder real-time prediction speed and accuracy. In this paper, we propose a novel Deep Learning architecture based on Convolutional Neural Networks (CNNs) for accurate classification of Arabic handwritten characters. Our approach offers simplicity without compromising accuracy, making it suitable for online recognition tasks. We validate our method on the Arabic Handwritten Characters Database (AHCD) and achieve a high recognition rate of 99%. The trained model demonstrates robust performance, indicating its potential for practical applications in Arabic character recognition.