A TEMPLATE CLASSIFIER ON BINARY MATRICES FOR HANDWRITTEN CHARACTER RECOGNITION OF THE KYRGYZ ALPHABET
Keywords:
pattern recognition, handwritten text, Kyrgyz alphabet, template method, Hamming distance, binary features, computational complexityAbstract
The paper addresses handwritten character recognition for the Kyrgyz alphabet, whose specific difficulty is caused by three letters — Ө, Ң and Ү — that differ from the graphically close О, Н and У only by a small local element. A template classifier operating on binary matrices of fixed size is formalised. It is shown that for binary vectors the decision rule based on the minimum Euclidean norm coincides with the rule based on the minimum Hamming distance, which allows arithmetic to be replaced by bitwise operations and reduces the cost of a single template comparison to a few machine words. A condition of stable discriminability for graphically close pairs is introduced, and a model estimate of the area of the distinguishing element is obtained as a function of grid resolution. The calculation shows that on a 10×10 grid the discriminability margin of the pair О — Ө is small, while an acceptable ratio is reached starting from a 28×28 grid. A two-stage scheme is described in which the per-character decision is refined by a dictionary check of the assembled word, together with the architecture of a web prototype based on Python, Django and PostgreSQL. The work is methodological: it provides a formal description of the method, complexity and discriminability estimates, and a programme of the measurement experiment required to verify these estimates quantitatively.
References
1. Журавлёв, Ю. И. Распознавание. Классификация. Прогноз. Математические методы и их применение / Ю. И. Журавлёв.– Москва: Наука, 1999. – Вып. 2.
2. Фисенко, В. Т. Компьютерная обработка и распознавание изображений: учебное пособие / В.Т.Фисенко. – Санкт-Петербург : СПбГУ ИТМО, 2018. – 192 с.
3. Фу, К. С. Лингвистический подход к распознаванию образов / К. С. Фу. – Москва : Мир, 2010.
4. Bishop, C. M. Pattern Recognition and Machine Learning [Text] / C. M. Bishop. – New York : Springer, 2006. – 738 p.
5. Duda, R. O. Pattern Classification [Text] / R. O. Duda, P. E. Hart, D. G. Stork. – 2nd ed. – New York: Wiley, 2001.
6. Dalal, N. Histograms of Oriented Gradients for Human Detection [Text] / N. Dalal, B. Triggs // Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition. – 2005. – Vol. 1. – P. 886–893.
7. Ojala, T. Multiresolution Gray-Scale and Rotation Invariant Texture Classification with Local Binary Patterns [Text] / T. Ojala, M. Pietikäinen, T.Mäenpää// IEEE Transactions on Pattern Analysis and Machine Intelligence. – 2002. – Vol. 24, No. 7. – P. 971–987.
8. LeCun, Y. Deep learning [Text] / Y. LeCun, Y. Bengio, G. Hinton // Nature. – 2015. – Vol. 521, No. 7553. – P. 436–444.
9. Simard, P. Y. Best Practices for Convolutional Neural Networks Applied to Visual Document Analysis [Text] / P. Y. Simard, D. Steinkraus, J. C. Platt // Proceedings of the Seventh International Conference on Document Analysis and Recognition. – 2003. – Vol. 3. – P. 958–962.
10. Cyrillic-MNIST: a Cyrillic Version of the MNIST Dataset [Text] / B. Tleubayev, A. Khassanov, R. Mussabayev [et al.] // Proceedings of the Thirteenth Language Resources and Evaluation Conference. – Marseille: European Language Resources Association, 2022.
11. Warren, H. S. Hacker’s Delight [Text] / H. S. Warren. – 2nd ed. – Boston: Addison-Wesley, 2012. – 512 p.
