HYBRID CNN-SVM APPROACH FOR UZBEK LATIN HANDWRITTEN CHARACTER RECOGNITION

Iskandarov S. . & Rokhmonova M. .


Annotatsiya

Handwritten character recognition for the Uzbek Latin alphabet remains an underexplored yet practically important problem due to the language's unique diacritical characters (o', g', sh, ch, ng). This paper proposes a hybrid deep learning approach that integrates Convolutional Neural Network (CNN) feature extraction with Support Vector Machine (SVM) classification to recognize Uzbek Latin handwritten characters. A dedicated dataset of 27,300 labeled samples was assembled from 155 writers. The preprocessing pipeline employs Sauvola adaptive binarization, geometric normalization, and a novel diacritical linking algorithm that recovers apostrophe marks misdetected as separate components. The CNN extracts a 128-dimensional feature vector fed directly into an RBF-kernel SVM, eliminating the standard Softmax head. Experimental results on the held-out test set yield 95.8% accuracy and an F1-score of 95.5%, surpassing standalone CNN (91.7%), HOG+SVM (81.4%), VGG-16 transfer learning (93.1%), and ResNet-50 transfer learning (94.2%) baselines. Ablation experiments confirm that the diacritical linking module alone contributes a +4.4% accuracy gain. The inference latency of 18 ms per character enables real-time deployment, and a TensorFlow Lite variant achieves 7 ms on a mobile Snapdragon processor with only 0.8% accuracy loss.


Adabiyotlar

  1. UNESCO, "A Guide to Best Practices for the Digitization of Handwritten Cultural Heritage," UNESCO Digital Library, Paris, France, Tech. Rep., 2021.
  2. Y. LeCun, L. Bottou, Y. Bengio, and P. Haffner, "Gradient-based learning applied to document recognition," Proc. IEEE, vol. 86, no. 11, pp. 2278–2324, Nov. 1998, doi: 10.1109/5.726791.
  3. A. Graves and J. Schmidhuber, "Offline Arabic handwriting recognition using multidimensional recurrent neural networks," in Guide to OCR for Arabic Scripts, Springer, 2012, pp. 297–313.
  4. G. Cohen, S. Afshar, J. Tapson, and A. van Schaik, "EMNIST: Extending MNIST to handwritten letters," in Proc. IJCNN, Anchorage, AK, May 2017, pp. 2921–2926, doi: 10.1109/IJCNN.2017.7966217.
  5. U.-V. Marti and H. Bunke, "The IAM-database: An English sentence database for offline handwriting recognition," Int. J. Doc. Anal. Recognit., vol. 5, no. 1, pp. 39–46, 2002, doi: 10.1007/s100320200071.
  6. N. Dalal and B. Triggs, "Histograms of oriented gradients for human detection," in Proc. IEEE CVPR, San Diego, CA, Jun. 2005, pp. 886–893, doi: 10.1109/CVPR.2005.177.
  7. I. Goodfellow, Y. Bengio, and A. Courville, Deep Learning. Cambridge, MA: MIT Press, 2016.
  8. X.-X. Niu and C. Y. Suen, "A novel hybrid CNN–SVM classifier for recognizing handwritten digits," Pattern Recognit., vol. 45, no. 4, pp. 1318–1325, Apr. 2012, doi: 10.1016/j.patcog.2011.09.021.
  9. S. Ahlawat and A. Choudhary, "Hybrid CNN-SVM classifier for handwritten digit recognition," Procedia Comput. Sci., vol. 167, pp. 2554–2560, 2020, doi: 10.1016/j.procs.2020.03.309.
  10. A. Graves, M. Liwicki, S. Fernández, R. Bertolami, H. Bunke, and J. Schmidhuber, "A novel connectionist system for unconstrained handwriting recognition," IEEE Trans. Pattern Anal. Mach. Intell., vol. 31, no. 5, pp. 855–868, May 2009, doi: 10.1109/TPAMI.2008.137.
  11. S. Peng, X. Liu, and C. Liu, "Attention-based Arabic handwriting recognition with multi-scale features," Pattern Recognit. Lett., vol. 138, pp. 610–617, 2020, doi: 10.1016/j.patrec.2020.08.022.
  12. I. Lozinskiy, A. Suleymanov, and B. Nurmukhanov, "Low-resource language models for Central Asian languages: Challenges and methods," in Proc. ACL Workshop Low-Resource Lang., 2023, pp. 112–119.
  13. S. Iskandarova and Z. Kuchkarov, "Recognition of Uzbek handwritten texts by selection of activation functions for convolutional neural network," Science and Innovation, vol. 2, no. A2, pp. 16–21, 2023.
  14. S. Iskandarov and M. Rakhmonova, "Theoretical foundations and efficient algorithms for feature detection and extraction," Raqamli Transformatsiya va Sun'iy Intellekt, vol. 3, no. 6, pp. 122–126, Dec. 2025, ISSN: 3030-3346.
  15. J. Sauvola and M. Pietikäinen, "Adaptive document image binarization," Pattern Recognit., vol. 33, no. 2, pp. 225–236, Feb. 2000, doi: 10.1016/S0031-3203(99)00055-2.
  16. P. Soille, Morphological Image Analysis: Principles and Applications, 2nd ed. Berlin: Springer, 2003.
  17. J. Liang, D. Doermann, and H. Li, "Page segmentation using a Voronoi+MST approach," in Proc. SPIE Document Recognit. Retrieval, vol. 5010, 2003, pp. 96–104.
  18. J. Cohen, "A coefficient of agreement for nominal scales," Educ. Psychol. Meas., vol. 20, no. 1, pp. 37–46, 1960, doi: 10.1177/001316446002000104.
  19. C. Shorten and T. M. Khoshgoftaar, "A survey on image data augmentation for deep learning," J. Big Data, vol. 6, no. 1, pp. 1–48, Jul. 2019, doi: 10.1186/s40537-019-0197-0.
  20. N. Otsu, "A threshold selection method from gray-level histograms," IEEE Trans. Syst., Man, Cybern., vol. 9, no. 1, pp. 62–66, Jan. 1979, doi: 10.1109/TSMC.1979.4310076.
  21. W. Niblack, An Introduction to Digital Image Processing. Englewood Cliffs, NJ: Prentice Hall, 1986.
  22. D. Bradley and G. Roth, "Adaptive thresholding using the integral image," J. Graph. Tools, vol. 12, no. 2, pp. 13–21, 2007, doi: 10.1080/2151237X.2007.10129236.
  23. S. Ioffe and C. Szegedy, "Batch normalization: Accelerating deep network training by reducing internal covariate shift," in Proc. ICML, Lille, France, 2015, vol. 37, pp. 448–456.
  24. M. Lin, Q. Chen, and S. Yan, "Network in network," in Proc. ICLR, Banff, Canada, 2014, arXiv:1312.4400.
  25. N. Srivastava, G. Hinton, A. Krizhevsky, I. Sutskever, and R. Salakhutdinov, "Dropout: A simple way to prevent neural networks from overfitting," J. Mach. Learn. Res., vol. 15, no. 1, pp. 1929–1958, 2014.
  26. D. P. Kingma and J. Ba, "Adam: A method for stochastic optimization," in Proc. ICLR, San Diego, CA, 2015, arXiv:1412.6980.
  27. M. Sokolova and G. Lapalme, "A systematic analysis of performance measures for classification tasks," Inf. Process. Manag., vol. 45, no. 4, pp. 427–437, 2009, doi: 10.1016/j.ipm.2009.03.002.
  28. D. Bhatt, C. Patel, H. Ghetiya, D. Pandya, and C. Bhatt, "Survey on various image classification techniques," J. Comput. Sci., vol. 17, no. 3, pp. 235–253, 2021.
  29. K. Simonyan and A. Zisserman, "Very deep convolutional networks for large-scale image recognition," in Proc. ICLR, San Diego, CA, 2015, arXiv:1409.1556.
  30. K. He, X. Zhang, S. Ren, and J. Sun, "Deep residual learning for image recognition," in Proc. IEEE CVPR, Las Vegas, NV, Jun. 2016, pp. 770–778, doi: 10.1109/CVPR.2016.90.
  31. A. G. Howard et al., "MobileNets: Efficient convolutional neural networks for mobile vision applications," arXiv:1704.04861, Apr. 2017.
  32. S. Ruder, "Neural transfer learning for natural language processing," Ph.D. thesis, Dept. Comput. Sci., Nat. Univ. Ireland, Galway, 2019.