Combining Handwriting Dynamics with ConvNeXtV2 Convolutional Backbone for Handwritten Character Recognition on Russian- and English-language Datasets
https://doi.org/10.15514/ISPRAS-2026-38(2)-13
Abstract
This work extends the author’s previous research on leveraging dynamic pen-motion characteristics to improve handwritten text recognition. We combine the earlier proposed dynamic component–angular and spectral trajectory features–with a modern visual backbone, ConvNeXtV2_tiny, and evaluate on three datasets: EMNIST (by_class), UJI Pen Characters 2, and our own Russian Handwritings Tracked. We show that, after moderate augmentation tuning, the visual branch achieves state-of-the-art performance on EMNIST among the models compared, while the dynamic branch and their ensemble improve robustness on datasets with greater handwriting variability (UJI, Russian Handwritings Tracked). The results confirm the complementarity of visual and kinematic features and highlight the promise of the method for “difficult handwriting” scenarios.
About the Author
Dmitry Vladimirovich IATSENKO
Institute of High Technologies and Piezotechnics, SFedU
Russian Federation
Russian Federation
Senior lecturer in the Department of Applied Informatics and Innovation at the Institute of High Technologies and Piezotechnics at SFedU since 2020. His research interests include machine learning, deep learning models, and cryptographic methods for information security.
References
1. Woo S., Debnath S., Hu R., Chen X., Liu Z., Kweon I. S., Xie S. Convnext v2: Codesigning and scaling convnets with masked autoencoders, in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2023, pp. 16133-16142. Available at: https://arxiv.org/pdf/2301.00808, accessed 10.01.2026.
2. Woo S. et al., Convnext v2: Co-designing and scaling convnets with masked autoencoders,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2023. Global Response Normalization (GRN)
3. Saenko I., Lauta O., Iatsenko D. The use of dynamic characteristics in handwriting recognition tasks, in 2023 International Ural Conference on Electrical Power Engineering (UralCon), 2023, pp. 609-614.
4. Яценко Д. В., Граецкая О. В., Смирнов К. В. Набор данных рукописных символов русского языка. Вестник компьютерных и информационных технологий, Feb. 2025, c. 46-54.
5. Prat F., Castro M., Llorens D., Marzal A., Vilar J. UJI Pen Characters (Version 2), 2008. Dataset is available at: https://doi.org/10.24432/C5FG8S, accessed 10.01.2026.
6. Llorens D., Prat F., Marzal A., Vilar J. M., Castro M. J., Amengual J. C., Barrachina S., Castellanos A., España S., Gómez J. A., Gorbe J., Gordo A., Palazón V., Peris G., Ramos-Garijo R., Zamora F. The UJIpenchars database: a pen-based database of isolated handwritten characters, in Proceedings of the Sixth International Conference on Language Resources and Evaluation (LREC’08), (Marrakech, Morocco), European Language Resources Association (ELRA), 2008.
7. Iatsenko D., Smirnov K. Russian handwritings tracked. Mendeley Data, V2, 2023. Dataset.
8. Hugging Face, Convnext v2 – transformers documentation, 2025. Available at: https://huggingface.co/docs/transformers/model_doc/convnextv2, accessed 04.10.2025.
9. Hastie T., Tibshirani R., Friedman J. The Elements of Statistical Learning: Data Mining, Inference, and Prediction. New York: Springer, 2 ed., 2009.
10. Geman S., Bienenstock E., Doursat R. Neural networks and the bias/variance dilemma, Neural Computation, vol. 4, no. 1, pp. 1-58, 1992.
11. LeCun Y., Bottou L., Bengio Y., Haffner P. Gradient-based learning applied to document recognition, Proceedings of the IEEE, vol. 86, no. 11, pp. 2278-2324, 1998.
12. Yacenko D. Article17: репозиторий GitHub (релиз v1.0) [Электронный ресурс]. URL: github.com/d-yacenko/Article17.git (дата обращения 10.01.2026)
13. Cohen G., Afshar S., Tapson J., van Schaik A. EMNIST: an extension of MNIST to handwritten letters, 2017. Available at: https://arxiv.org/pdf/1702.05373, accessed 10.01.2026.
Review
For citations:
IATSENKO D.V. Combining Handwriting Dynamics with ConvNeXtV2 Convolutional Backbone for Handwritten Character Recognition on Russian- and English-language Datasets. Proceedings of the Institute for System Programming of the RAS (Proceedings of ISP RAS). 2026;38(2):195-203. (In Russ.) https://doi.org/10.15514/ISPRAS-2026-38(2)-13
Email this article 