Elamur Research⌕
☰
home / papers / 2020 / Neural Computing for Online Arabic Handwriting Character Recognition using Hard Stroke Features Mining

Neural Computing for Online Arabic Handwriting Character Recognition using Hard Stroke Features Mining

Summary

The paper tackles the recognition of isolated online Arabic handwritten characters. Variable styles, writing speed and small diacritic differences between letters keep accuracy below what Latin and Chinese systems reach. The pipeline has four stages:

  • Smoothing: pen trajectories are cleaned with a neighbour-weighted interpolation filter.
  • Segmentation: each stroke is cut into smaller units called tokens. The method first decides whether a stroke is mainly horizontal or vertical by comparing its x-extent and y-extent. It then finds critical points: local maxima backed by monotone runs of points on both sides, with run length tied to the size of the pattern.
  • Features: a small set of structural and statistical descriptors is computed per token and binarised. They include orientation, length, direction, midpoint and length relative to the parent stroke.
  • Classification: a backpropagation multilayer perceptron with a modified sigmoid activation assigns the character class.

A stated design goal is to keep the number of tokens per character small and stable, so the classifier gets a fixed-size input. On characters from the OHASD benchmark, the system reports 98.6% average accuracy. This is competitive with earlier Arabic recognisers and suggests that careful segmentation can reduce the need for large feature sets.

Key points

  • Data and result: the dataset is OHASD, split 70% for training and 30% for testing, with a reported average accuracy of 98.6%.
  • Classifier settings:
    • BP/MLP with a modified sigmoid activation;
    • initial weights drawn from (−0.2, 0.2);
    • learning rate 0.1;
    • momentum 0.05;
    • internal threshold 0.0.
  • Critical points: a point counts as a local maximum only if y-values rise and fall monotonically over a window of about 0.05×N points on each side.
  • Token-count stability (Table 5): the minimum token count occurs for 69% of characters, the most frequent count for 74%, and the maximum for 11%. Only Alef, Jeem, Ain and Yeh deviate from this pattern.
  • Remaining errors: most come from look-alike pairs, notably initial-form Sad versus Seen. Delayed secondary strokes help separate pairs such as Seen/Sheen and Ain/Ghain.
  • Earlier results for comparison:
    • zoning plus EBPANN: 93.61%;
    • rule-based projection profiles: 97.6%;
    • ANN–PSO on the TMU dataset: 88.47%;
    • fuzzy ARTMAP on IFN/ENIT: 94.72%.

Interesting ideas

  • I1. Before choosing cut points, decide each unit's dominant axis by comparing its horizontal and vertical extents, then search for boundaries only along that axis; this adapts the segmentation rule to the shape of each unit instead of using one global rule.
  • I2. A boundary is accepted only when supported by a consistent monotone run on both sides, with run length proportional to input size; this filters jitter and very short breaks without hand-tuned absolute thresholds.
  • I3. The stability of a decomposition is measured as how often a character yields its most frequent token count, and that stability is treated as a goal in itself.
  • I4. Features are expressed as ratios to the parent unit (token length as a percentage of stroke length), making them invariant to writing size.
  • I5. Small detached marks are kept as separate delayed strokes but used together with the main body to separate near-identical letters, which is where the remaining errors concentrate.

For Mova

How these ideas could serve Mova, our small open language model — its files, stages and tests:

  • I1 (tokens): the tokenizer could first classify a span's type (number, URL, hyphenated compound, ordinary word) and then apply a boundary rule specific to that type, logging which rule fired so each split stays explainable.
  • I2 (learn): when mining self-induced repair rules, accept a candidate only if its evidence persists across a window of examples scaled to corpus size, which guards against rules triggered by a single noisy parse.
  • I3 (learn): for silver data from several witness parsers, track how often each construction yields its modal number of nodes or frames, and flag low-stability constructions for teacher review.
  • I4 (tree): parser features such as dependency distance could be normalised by sentence or clause length, so behaviour transfers between short school texts and long sentences.
  • I5 (tokens): treat clitics, apostrophes and hyphens as attached secondary cues to the host word (e.g. 's vs s', it's vs its), and run error analysis on such confusable pairs to show which cue decided each tag.

Our summary of the paper, in our own words — not its text. The work belongs to its authors; read it at the links above.

Fields: Linguistics & languages

Topics Semitic languages

Cite

Cite as Amjad Rehman. 2020. Neural Computing for Online Arabic Handwriting Character Recognition using Hard Stroke Features Mining. arXiv:2005.02171 · IJICIC 2021.

The paper itself, not this page. BibTeX from the library's record of the work. DOI 10.48550/arxiv.2005.02171 · arXiv 2005.02171

@misc{rehman2020neural,
  title = {{Neural Computing for Online Arabic Handwriting Character Recognition using Hard Stroke Features Mining}},
  author = {Amjad Rehman},
  year = {2020},
  doi = {10.48550/arxiv.2005.02171},
  eprint = {2005.02171},
  archivePrefix = {arXiv},
  url = {https://doi.org/10.48550/arxiv.2005.02171}
}

References

70 references, 0 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. T. Saba, A. Rehman, and G. Sulong, "Improved statistical features for cursive character recognition," International Journal of Innovative Computing, Information and Control, vol. 7, pp. 5211-5224, 2011. find
  2. T. Saba, A. Rehman, A. Altameem, and M. Uddin, "Annotated comparisons of proposed preprocessing Most repetitive tokens 74% Maximum 11% (23) techniques for script recognition," Neural Computing and Applications, vol. 25, pp. 1337-1347, 2014. find
  3. M. Sharif, M. A. Khan, M. Faisal, M. Yasmin, and S. L. Fernandes, "A framework for offline signature verification system: Best features selection approach," Pattern Recognition Letters, 2018. find
  4. T. Saba, A. S. Almazyad, and A. Rehman, "Online versus offline Arabic script classification," Neural Computing and Applications, vol. 27, pp. 1797-1804, 2016. find
  5. Ebrahim, A.Y., Kolivand, H., Rehman, A., Rahim, M.S.M. and Saba, T. (2018) ‘Features selection for offline handwritten signature verification: state of the art’, Int. J. Computational Vision and Robotics, vol. 8(6), pp.606–622 [5]T. Saba and F. A. Alqahtani, "Semantic analysis based forms information retrieval and classification," 3D Research, vol. 4, p. 4, 2013. find
  6. Haji, M.S. Alkawaz, M.H. Rehman, A. Saba, T. (2019) Content-based image retrieval: a deep look at features prospectus, International Journal of Computational Vision and Robotics 9 (1), 14-38 find
  7. Z. Muhsin, A. Rehman, A. Altameem, T. Saba, and M. Uddin, "Improved quadtree image segmentation approach to region information," The Imaging Science Journal, vol. 62, pp. 56-62, 2014. find
  8. M. S. Fadhil, M. H. Alkawaz, A. Rehman, and T. Saba, "Writers identification based on multiple windows features mining," 3D Research, vol. 7, p. 8, 2016. find
  9. A. Rehman, F. Kurniawan, and T. Saba, "An automatic approach for line detection and removal without smash-up characters," The Imaging Science Journal, vol. 59, pp. 177-182, 2011. find
  10. Ebrahim, A.Y., Kolivand, H., Rehman, A., Rahim, M.S.M. and Saba, T. (2018) ‘Features selection for offline handwritten signature verification: state of the art’, Int. J. Computational Vision and Robotics, vol. 8(6), pp.606–622 find
  11. E. F. B. Tasdemir and B. Yanikoglu, "A comparative study of delayed stroke handling approaches in online handwriting," International Journal on Document Analysis and Recognition (IJDAR), vol. 22, pp. 15-28, 2019. find
  12. S. Miyagawa, K. Bulert, M. Büchler, and H. Behlmer, "Optical character recognition of typeset Coptic text with neural networks," Digital Scholarship in the Humanities, 2019. find
  13. Alkawaz,M.H., Sulong, G. Saba, T. Almazyad, A.S. Rehman, A. (2017) Concise analysis of current text automation and watermarking approaches, Security and Communication Networks 9 (18), 6365-6378 find
  14. MA Khan, M Sharif, MY Javed, T Akram, M Yasmin, T Saba License number plate recognition system using entropy-based features selection approach with SVM, IET Image Processing 12 (2), 200-209 (2017) find
  15. Mehmood, Z., Abbas, F., Mahmood, T. et al. Content-Based Image Retrieval Based on Visual Words Fusion Versus Features Fusion of Local and Global Features, Arab J Sci Eng (2018). https://doi.org/10.1007/s13369-018-3062-0. ↗
  16. U. Sharif, Z. Mehmood, T. Mahmood, M. A. Javid, A. Rehman, and T. Saba, "Scene analysis and search using local features and support vector machine for effective content-based image retrieval," Artificial Intelligence Review, pp. 1-25, 2018. find
  17. S. Jabeen, Z. Mehmood, T. Mahmood, T. Saba, A. Rehman, and M. T. Mahmood, "An effective content-based image retrieval technique for image visuals representation based on the bag-of-visual-words model," PloS one, vol. 13, p. e0194526, 2018. find
  18. Rehman, A., Saba, T., Mahmood, T., Mehmood, Z., Shah,M., Anjum, A. (2018) Data hiding technique in steganography for information security using number theory, Journal of Information Science, 0165551518816303 find
  19. M. Sharif, M. A. Khan, T. Akram, M. Y. Javed, T. Saba, and A. Rehman, "A framework of human detection and action recognition based on uniform segmentation and combination of Euclidean distance and joint entropy-based features selection," EURASIP Journal on Image and Video Processing, vol. 2017, p. 89, 2017. find
  20. S. A. Khan, M. Nazir, M. A. Khan, T. Saba, K. Javed, A. Rehman, et al., "Lungs nodule detection framework from computed tomography images using support vector machine," Microscopy research and technique, 2019. find
  21. M. A. Khan, T. Akram, M. Sharif, T. Saba, K. Javed, I. U. Lali, et al., "Construction of saliency map and hybrid set of features for efficient segmentation and classification of skin lesion," Microscopy research and technique, 2019. find
  22. Meethongjan,K., Dzulkifli,M., Rehman,A., Altameem,A., Saba, T. (2013) An intelligent fused approach for face recognition, Journal of Intelligent Systems vol.22(2), pp. 197-212. find
  23. Iqbal, S. Ghani, M.U. Saba, T. Rehman, A. (2018). Brain tumor segmentation in multi-spectral MRI using convolutional neural networks (CNN). Microsc Res Tech. 81(4):419-427. doi: 10.1002/jemt.22994. ↗
  24. Khan, MA.,Sharif, M. Akram, T., Raza, M., Saba,T., Rehman, A. (2020) Hand-crafted and deep convolutional neural network features fusion and selection strategy: an application to intelligent human action recognition, Applied Soft Computing 87, 105986 find
  25. S. Saeed, S. Naz, and M. I. Razzak, "An Application of Deep Learning in Character Recognition: An Overview," in Handbook of Deep Learning Applications, ed: Springer, 2019, pp. 53-81. find
  26. Jianhua Zhang, Zhiqiang Wu, Guanglei Cui, Penghao Fan, Jijun Cao and Mingyue Ning, Sensor Fault Diagnosis Based on Correntropy Filter and Probabilistic Neural Network, ICIC Express Letters, vol.13, no.8, pp.743-751, 2019 find
  27. Abed M. and Alasad H., “High Accuracy Arabic Handwritten Characters Recognition Using Error Back Propagation Artificial Neural Networks,” International Journal of Advanced Computer Science and Applications, vol. 6, no. 2, pp. 145-152, 2015. find
  28. M. Harouni, D. Mohamad, M. S. M. Rahim, S. M. Halawani, and M. Afzali, "Handwritten Arabic character recognition based on minimal geometric features," International Journal of Machine Learning and Computing, vol. 2, p. 578, 2012. find
  29. M. Harouni, M. Rahim, M. Al-Rodhaan, T. Saba, A. Rehman, and A. Al-Dhelaan, "Online Persian/Arabic script classification without contextual information," The Imaging Science Journal, vol. 62, pp. 437-448, 2014. find
  30. Ramzi, A., and A. Zahary. 2014. Online Arabic handwritten character recognition using online-offline feature extraction and back-propagation neural network. In First international conference on advanced technologies for signal and image processing (ATSIP), 350–55. IEEE/ATSIP. doi:10.1109/ATSIP.2014.6834634 ↗
  31. Baligh M. Al-Helali & Sabri A. Mahmoud (2016) A Statistical Framework for Online Arabic Character Recognition, Cybernetics and Systems, 47:6, 478-498 find
  32. T. Saba, "Fuzzy ARTMAP Approach for Arabic Writer Identification using Novel Features Fusion," Journal of Computer Science, vol. 14, pp. 210-220, 2018. find
  33. M. A. Khan, I. U. Lali, A. Rehman, M. Ishaq, M. Sharif, T. Saba, et al., "Brain tumor detection and classification: A framework of marker‐based watershed algorithm and multilevel priority features selection," Microscopy research and technique, 2019. find
  34. F Ramzan, MUG Khan, A Rehmat, S Iqbal, T Saba, A Rehman, Z Mehmood (2020) A Deep Learning Approach for Automated Diagnosis and Multi-Class Classification of Alzheimer’s Disease Stages Using Resting-State fMRI and Residual Neural Networks, Journal of Medical Systems 44 (2), 37 find
  35. Saba,T. Almazyad, A.S., Rehman, A. (2015) Language independent rule based classification of printed & handwritten text, IEEE International Conference on Evolving and Adaptive Intelligent Systems (EAIS), pp. 1-4, doi. 10.1109/EAIS.2015.7368806. ↗
  36. Iqbal, S. Ghani, M.U. Saba, T. Rehman, A. (2018). Brain tumor segmentation in multi-spectral MRI using convolutional neural networks (CNN). Microsc Res Tech. 81(4):419-427. doi: 10.1002/jemt.22994. ↗
  37. Sharif, M. Khan, M.A Akram, T. Javed, M.Y. Saba,T. Rehman, A (2017) A framework of human detection and action recognition based on uniform segmentation and combination of Euclidean distance and joint entropy-based features selection, EURASIP Journal on Image and Video Processing 2017 (1), 89, pp.1-18. find
  38. Saba, T. Rehman, A. Sulong, G. (2010). Non-linear segmentation of touched roman characters based on genetic algorithm, International Journal of Computer Science and Engineering, vol 2(6),pp. 2167-2172 find
  39. Muhsin; Z.F. Rehman, A.; Altameem, A.; Saba, A.; Uddin, M. (2014). Improved quadtree image segmentation approach to region information. the imaging science journal, vol. 62(1), pp. 56-62, doi. http://dx.doi.org/10.1179/1743131X13Y.0000000063. ↗
  40. Saba T, Bokhari STF, Sharif M, Yasmin M, Raza M. (2018b) Fundus image classification methods for the detection of glaucoma: A review, Microsc Res Tech. doi: 10.1002/jemt.23094. ↗
  41. Sharif,U., Mehmood, Z. Mahmood, T., Javid, M.A. Rehman,A. Saba, T. (2018) Scene analysis and search using local features and support vector machine for effective content-based image retrieval, Artificial Intelligence Review, pp. 1-25 find
  42. Khan S.A, Nazir M, Khan M.A, Saba T, Javed K, Rehman A, Akram T, Awais M ... (2019).Lungs nodule detection framework from computed tomography images using support vector machine, Microsc Res Tech. doi: 10.1002/jemt.23275 ↗
  43. Saba, T., Khan, S.U., Islam, N., Abbas, N., Rehman, A., Javaid, N. and Anjum, A., 2019. Cloud‐based decision support system for the detection and classification of malignant cells in breast cancer using breast cytology images. Microscopy research and technique, 82(6), pp.775-785. find
  44. T Saba (2012) Offline Cursive Touched Script Non-linear Segmentation Universiti Teknologi Malaysia, Malaysia, PhD thesis PP. 33-39. find
  45. K. Neamah, D. Mohamad, T. Saba, and A. Rehman, "Discriminative features mining for offline handwritten signature verification," 3D Research, vol. 5, p. 2, 2014. find
  46. Saba, T., and A. Rehman. "Machine learning and script recognition." (2012)LAP LAMBERT Academic Publishing: pp. 37-39. find
  47. Nodehi, A., Sulong, G., Al-Rodhaan, M., Al-Dhelaan, A., Rehman, A. and Saba, T., 2014. Intelligent fuzzy approach for fast fractal image compression. EURASIP Journal on Advances in Signal Processing, 2014(1), p.112, doi. 10.1186/1687-6180-2014-112. ↗
  48. Saba,T., Rehman, A., Al-Dhelaan, A., Al-Rodhaan, M. (2014) Evaluation of current documents image denoising techniques: a comparative study , Applied Artificial Intelligence, vol.28 (9), pp. 879-887, doi. 10.1080/08839514.2014.954344 ↗
  49. Seni, G. Anastasakos, T. "Non-cumulative character scoring in a forward search for Online handwriting recognition", Proceedings of the IEEE International Conference on Engineering Acoustics, Speech, and Signal Processing, 2000. p.3450-3453. find
  50. Haron,H. Rehman, A. Wulandhari, L.A. Saba, T. (2011). Improved vertex chain code based mapping algorithm for curve length estimation, Journal of Computer Science, vol. 7(5), pp. 736-743, doi. 10.3844/jcssp.2011.736.743. ↗
  51. Saba T. (2019) Automated lung nodule detection and classification based on multiple classifiers voting. Microsc Res Tech. 2019;1–9. doi.org/10.1002/jemt.23326 ↗
  52. Jadooki, S. Mohamad,D., Saba, T., Almazyad, A.S. Rehman, A. (2016) Fused features mining for depth-based hand gesture recognition to classify blind human communication, Neural Computing and Applications, Vol.28(11), pp 3285–3294, doi. 10.1007/s00521-016-2244-5 ↗
  53. Rehman, A., Abbas, N., Saba, T., Rahman, S.I.U., Mehmood, Z. and Kolivand, H., 2018. Classification of acute lymphoblastic leukemia using deep learning. Microscopy Research and Technique, 81(11), pp.1310-1317 find
  54. Mughal, B. Muhammad, N. Sharif, M. Rehman, A. Saba, T. (2018) Removal of pectoral muscle based on topographic map and shape-shifting silhouette, BMC Cancer, https://doi.org/10.1186/s12885-018-4638-5 ↗
  55. Sharif, M., Naz, F., Yasmin, M., Shahid, M.A. and Rehman, A., 2017. Face Recognition: A Survey. Journal of Engineering Science & Technology Review, 10(2) find
  56. Saba, T., Rehman,A. Mehmood, Z., Kolivand, H., Sharif, M.(2018a) Image Enhancement and Segmentation Techniques for Detection of Knee Joint Diseases: A Survey, Current Medical Imaging Reviews, vol. 14(5), doi. 10.2174/1573405613666170912164546. ↗
  57. Mughal, B. Muhammad, N. Sharif, M. Rehman, A. Saba, T. (2018) Removal of pectoral muscle based on topographic map and shape-shifting silhouette, BMC Cancer, https://doi.org/10.1186/s12885-018-4638-5 ↗
  58. Ullah,H., Saba,T. Islam, N. Abbas, N. Rehman, A., Mehmood, Z. Anjum, A. (2019) An ensemble classification of exudates in color fundus images using an evolutionary algorithm based optimal features selection, Microscopy research and technique, doi.org/10.1002/jemt.23178 ↗
  59. Rehman, A. Abbas, N. Saba, T. Mehmood, Z. Mahmood, T. Ahmed, K.T. (2018b) Microscopic malaria parasitemia diagnosis and grading on benchmark datasets, Microscopic Research and Technique, 81(9):1042-1058. doi: 10.1002/jemt.23071 ↗
  60. Iqbal, S., Khan,M.U.G., Saba, T. Mehmood, Z. Javaid,N., Rehman,A., Abbasi, R. (2019) Deep learning model integrating features and novel classifiers fusion for brain tumor segmentation, Microscopy Research and Technique, doi.org/10.1002/jemt.23281 ↗
  61. Randa Ibrahim Elanwar, Mohsen Rashwan, and Samia Mashali. 2010. OHASD: The first on-line arabic sentence database handwritten on tablet PC. In Proceedings of the World Academy of Science, Engineering and Technology (WASET’10) International Conference on International Conference on Signal and Image Processing (ICSIP’10). Singapore, 910–915. find
  62. Iqbal, S., Khan, M. U. G., Saba, T., Rehman, A. (2017). Computer assisted brain tumor type discrimination using magnetic resonance imaging features. Biomedical Engineering Letters, vol.8(1),pp.5-28, doi. 10.1007/s13534-017-0050-3. ↗
  63. Sharif, M., Naz, F., Yasmin, M., Shahid, M.A. and Rehman, A., 2017. Face Recognition: A Survey. Journal of Engineering Science & Technology Review, 10(2). find
  64. Saba, T., Rehman, A. and AlGhamdi, J.S., 2017. Weather forecasting based on hybrid neural model. Applied Water Science, 7(7), pp.3869-3874. find
  65. Saba, T., Khan, M.A., Rehman, A. and Marie-Sainte, S.L., 2019. Region extraction and classification of skin cancer: A heterogeneous framework of deep CNN features fusion and reduction. Journal of medical systems, 43(9), p.289. find
  66. MA Khan, M Sharif, MY Javed, T Akram, M Yasmin, T Saba License number plate recognition system using entropy-based features selection approach with SVM, IET Image Processing 12 (2), 200-209 (2017) find
  67. Al-Dabbagh, M.M., Salim, N., Rehman, A., Alkawaz, M.H., Saba, T., Al-Rodhaan, M. and Al-Dhelaan, A., 2014. Intelligent bar chart plagiarism detection in documents. The Scientific World Journal, 2014. find
  68. M Bilal, HA Habib, Z Mehmood, RM Yousaf, T Saba, A Rehman (2020) A robust technique for copy-move forgery detection from small and extremely smooth tampered regions based on the DHE-SURF features and mDBSCAN clustering Australian Journal of Forensic Sciences, 1-24 find
  69. Saba, T. Mohamed,AS. El-Affendi,M. Amin, J. Sharif, M (2020) Brain tumor detection using fusion of hand crafted and deep learning features Cognitive Systems Research 59, 221-230, find
  70. A Mittal, D Kumar, M Mittal, T Saba, I Abunadi, A Rehman, S Roy (2020) Detecting Pneumonia using Convolutions and Dynamic Capsule Routing for Chest X-ray Images Sensors 20 (4), 1068 find

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/rehman-2020-neural-computing-online-arabic-handwriting/