An Efficient Hybrid CRNN Framework with CTC for Offline Arabic Handwritten Text Recognition
AUTHORS
Youssef El Mahdaoui,Department of Computer Science, Faculty of Sciences and Techniques, Cadi Ayyad University, Marrakech, Morocco
Salma Benjelloun,Laboratory of Intelligent Systems and Applications, Mohammed V University in Rabat, Rabat, Morocco
ABSTRACT
Handwritten Text Recognition (HTR) for Arabic script remains a challenging problem due to its cursive nature, contextual character variations, and the presence of ligatures and diacritics. These characteristics complicate segmentation and sequence modeling, limiting the effectiveness of traditional recognition approaches. This paper proposes an efficient hybrid deep learning framework for offline Arabic handwritten text recognition that integrates Convolutional Neural Networks (CNNs), Bidirectional Long Short-Term Memory (BLSTMs), and Connectionist Temporal Classification (CTC). The proposed architecture enables end-to-end learning without explicit segmentation, allowing the model to map input text images to corresponding character sequences directly. To address data scarcity and improve generalization, a targeted data augmentation strategy is incorporated during training. The model is evaluated on two widely used benchmark datasets, IFN/ENIT and KHATT, representing word-level and line-level recognition tasks, respectively. Experimental results demonstrate that the proposed approach achieves competitive and state-of-the-art performance, with significant improvements in Word Error Rate (WER) and Character Error Rate (CER) compared to existing methods. The findings highlight the effectiveness of combining convolutional feature extraction with bidirectional sequence modeling for capturing the structural and contextual complexities of Arabic handwriting. Furthermore, the proposed system exhibits strong robustness across diverse writing styles and input conditions. This work provides a scalable, adaptable solution for real-world applications, including document digitization, archival processing, and intelligent text recognition systems. Future research directions include integrating attention mechanisms, advanced language modeling, and domain adaptation techniques to enhance performance and applicability further.
KEYWORDS
Arabic handwriting recognition, Deep learning, Convolutional Neural Networks (CNN), Bidirectional Long Short-Term Memory (BLSTM), Connectionist Temporal Classification (CTC), Optical Character Recognition (OCR), Sequence modeling, Data augmentation
REFERENCES
[1] M. Al-Ayyoub, A. Nuseir, K. Alsmearat, Y. Jararweh, and B. Gupta, “Deep learning for Arabic NLP: A survey,” Journal of Computational Science, vol.26, pp.522-531, (2018) DOI:10.1016/j.jocs.2017.11.011(CrossRef)(Google Scholar)
[2] H. Alsurori, A. A. Mohsen, G. Al-Badani, M. Al-Nahari, T. Faisal, and S. Alkasem, “Review on Arabic Handwritten Recognition Using Deep Learning and Machine Learning,” in Proceedings of the 3rd International Conference on Emerging Smart Technologies and Applications (eSmarTA), Taiz, Yemen, pp.1-8, (2023) DOI:10.1109/eSmarTA59349.2023.10293409(CrossRef)(Google Scholar)
[3] R. Najam and S. Faizullah, “A scarce dataset for ancient Arabic handwritten text recognition,” Data in Brief, vol.56, 110813, (2024) DOI:10.1016/j.dib.2024.110813(CrossRef)(Google Scholar)
[4] A. M. Mutawa, M. Y. Allaho, and M. Al-Hajeri, “Machine Learning Approach for Arabic Handwritten Recognition,” Applied Sciences, vol.14, no.19, 9020, (2023) DOI:10.3390/app14199020(CrossRef)(Google Scholar)
[5] R. C. Karpagalakshmi, S. Hegde, R. S. R, S. Mahapatra, E. S. Leni, and A. Sungheetha, “Deep Learning-Based Recognition of Handwritten English Characters,” Procedia Computer Science, vol.258, pp.1783-1792, (2024) DOI:10.1016/j.procs.2025.04.430(CrossRef)(Google Scholar)
[6] F. Alwajih, E. Badr, and S. Abdou, “Writer adaptation for E2E Arabic online handwriting recognition via adversarial multi-task learning,” Egyptian Informatics Journal, vol.23, no.3, pp.373-382, (2022) DOI:10.1016/j.eij.2022.02.007(CrossRef)(Google Scholar)
[7] N. Alzrrog, J.-F. Bousquet, and I. El-Feghi, “Deep Learning Application for Handwritten Arabic Word Recognition,” in Proceedings of the IEEE Canadian Conference on Electrical and Computer Engineering (CCECE), Halifax, NS, Canada, pp.95-100, (2022) DOI:10.1109/CCECE49351.2022.9918375(CrossRef)(Google Scholar)
[8] R. Ahmed, M. Gogate, A. Tahir, K. Dashtipour, B. Al-tamimi, A. Hawalah, M. A. El-Affendi, and A. Hussain, “Novel Deep Convolutional Neural Network-Based Contextual Recognition of Arabic Handwritten Scripts,” Entropy, vol.23, no.3, 340, (2021) DOI:10.3390/e23030340(CrossRef)(Google Scholar)
[9] I. Ahmad and G. A. Fink, “Handwritten Arabic text recognition using multi-stage sub-core-shape HMMs,” International Journal on Document Analysis and Recognition (IJDAR), vol.22, pp.329-349, (2019) DOI:10.1007/s10032-019-00339-8(CrossRef)(Google Scholar)
[10] L. Berriche, A. Alqahtani, and S. Rekik, “Hybrid Arabic handwritten character segmentation using CNN and graph theory algorithm,” Journal of King Saud University - Computer and Information Sciences, vol.36, no.1, 101872, (2023) DOI:10.1016/j.jksuci.2023.101872(CrossRef)(Google Scholar)
[11] B. Hakim and B. Ahror, “Deep Learning for Accurate Recognition of Arabic Handwritten Words in Historical Documents,” Procedia Computer Science, vol.244, pp.57-65, (2023) DOI:10.1016/j.procs.2024.10.178(CrossRef)(Google Scholar)
[12] R. Maalej and M. Kherallah, “New MDLSTM-based designs with data augmentation for offline Arabic handwriting recognition,” Multimedia Tools and Applications, vol.81, no.7, pp.1-20, (2022) DOI:10.1007/s11042-022-12339-8(CrossRef)(Google Scholar)
[13] S. R. Agre, C. G. Balaji, A. Salunke, O. Bagade, S. Prakash, and S. Goyal, “A Survey of Handwritten Text Recognition Approaches for Low-Resource Indic Scripts,” in Proceedings of the IEEE Pune Section International Conference (PuneCon), Pune, India, pp.1-5, (2025) DOI:10.1109/PuneCon67554.2025.11378784(CrossRef)(Google Scholar)
[14] T. Ben Aïcha Gader and A. Kacem Echi, “Attention-based deep learning model for Arabic handwritten text recognition,” Machine Graphics & Vision, vol.31, no.1/4, pp.49-73, (2022) DOI:10.22630/MGV.2022.31.1.3(CrossRef)(Google Scholar)
[15] A. Praneetha, S. Harisa, K. Satish, and B. Vivekananda, “Handwritten Text Recognition Using Word Beam Search,” in Proceedings of the 9th International Conference on Advanced Computing and Communication Systems (ICACCS), Coimbatore, India, pp.1878-1882, (2023) DOI:10.1109/ICACCS57279.2023.10112967(CrossRef)(Google Scholar)
[16] S. A. Mahmoud, I. Ahmad, W. G. Al-Khatib, M. Alshayeb, M. Tanvir Parvez, V. Märgner, and G. A. Fink, “KHATT: An open Arabic offline handwritten text database,” Pattern Recognition, vol.47, no.3, pp.1096-1112, (2014) DOI:10.1016/j.patcog.2013.08.009(CrossRef)(Google Scholar)
[17] C. Shorten and T. M. Khoshgoftaar, “A survey on Image Data Augmentation for Deep Learning,” Journal of Big Data, vol.6, no.60, pp.1-48, (2019) DOI:10.1186/s40537-019-0197-0(CrossRef)(Google Scholar)
[18] R. Najam and S. Faizullah, “A scarce dataset for ancient Arabic handwritten text recognition,” Data in Brief, vol.56, 110813, (2024) DOI:10.1016/j.dib.2024.110813(CrossRef)(Google Scholar)