정보과학회 컴퓨팅의 실제 논문지 (KIISE Transactions on Computing Practices)

  • 제23권5호
  • /
  • Pages.316-321
  • /
  • 2017
  • /
  • 2383-6318(pISSN)
  • /
  • 2383-6326(eISSN)
한국정보과학회 (Korean Institute of Information Scientists and Engineers)
권호 표지
DOI QR코드

DOI QR Code

순환신경망을 이용한 한글 필기체 인식

Hangul Handwriting Recognition using Recurrent Neural Networks

  • 투고 : 2016.10.10
  • 심사 : 2017.02.14
  • 발행 : 2017.05.15
⟨ 이전 논문 다음 논문 ⟩

초록

온라인 방식의 한글 필기체 인식 문제를 분석하고 순환신경망 기반의 해법을 모색한다. 한글 낱글자 인식 문제를 순서데이터 레이블링의 관점에서 서열 분류, 구간 분류, 시간별 분류의 세 단계로 구분하여 각각에 대한 해법을 살펴보며, 한글의 구성 원리를 고려한 해결 방안을 정리한다. 한글 2350글자에 대한 온라인 필기체 데이터에 GRU(gated recurrent unit)의 다층 구조를 가지는 서열 분류모델을 적용한 결과, 낱글자 인식 정확도는 86.2%, 초 중 종성 구성에 따른 6가지 유형 분류 정확도는 98.2%로 측정되었다. 유형 분류 모델로 획의 진행에 따른 유형 변화 역시 높은 정확도로 인식하는 결과를 통해, 순환신경망을 이용하여 순서 데이터에서 한글의 구조와 같은 고차원적 지식을 학습할 수 있음을 확인하였다.

We analyze the online Hangul handwriting recognition problem (HHR) and present solutions based on recurrent neural networks. The solutions are organized according to the three kinds of sequence labeling problem - sequence classifications, segment classification, and temporal classification, with additional consideration of the structural constitution of Hangul characters. We present a stacked gated recurrent unit (GRU) based model as the natural HHR solution in the sequence classification level. The proposed model shows 86.2% accuracy for recognizing 2350 Hangul characters and 98.2% accuracy for recognizing the six types of Hangul characters. We show that the type recognizing model successfully follows the type change as strokes are sequentially written. These results show the potential for RNN models to learn high-level structural information from sequential data.

키워드

과제정보

연구 과제 주관 기관 : 정보통신기술진흥센터, 한국산업기술평가관리원

참고문헌

  1. D. Ciresan, U. Meier, J. Schmidhuber, Multi-column Deep Neural Networks for Image Classification, Proc. CVPR, pp. 3642-3649, 2012.
  2. M. Liwicki, A. Graves, S. Fernández, H. Bunke, J. Schmidhuber, A Novel Approach to On-Line Handwriting Recognition Based on Bidirectional Long Short-Term Memory Networks, Proc. ICDAR, pp. 367-371, 2007.
  3. F. Yin, Q.-F. Wang, X.-Y. Zhang, and C.-L. Liu, ICDAR 2013 Chinese Handwriting Recognition Competition, Proc. ICDAR, pp. 1464-1470, 2013.
  4. A. Graves and J. J. Schmidhuber, Offline Handwriting Recognition with Multidimensional Recurrent Neural Networks, Proc. NIPS, pp. 545-552, 2008.
  5. B.-K. Sin and J. Kim, "Ligature Modeling for Online Cursive Script Recognition," IEEE Trans. Pattern Anal. Mach. Intell, Vol. 19, No. 6, pp. 623- 633, 1997. https://doi.org/10.1109/34.601250
  6. S. Cho and J. Kim, Bayesian Network Modeling of Hangul Characters for On-line Handwriting Recognition, Proc. ICDAR, pp. 207-211, 2003.
  7. I. J. Kim and X. H. Xie, "Handwritten Hangul Recognition Using Deep Convolutional Neural Networks," Int. J. Doc. Anal. Recognit., Vol. 18, No. 1, pp. 1-13, 2015. https://doi.org/10.1007/s10032-014-0229-4
  8. I. J. Kim, C. B Choi, and S. H. Lee, "Improving Discrimination Ability of Convolutional Neural Networks by Hybrid Learning," Int. J. Doc. Anal. Recognit., Vol. 19, No. 1, pp. 1-9, 2016. https://doi.org/10.1007/s10032-015-0256-9
  9. A. Graves, Supervised Sequence Labelling with Recurrent Neural Networks, Textbook, Studies in Computational Intelligence, Springer, 2012.
  10. H. Yoon, D. Jeong, and K. Jung, "Structure of Recurrent Neural Networks and Handwriting Recognition," Communications of the Korean Institute of Information Scientists and Engineers, Vol. 33, No. 9, pp. 42-48, 2015. (in Korean)
  11. J. L. Elman, "Finding Structure in Time," Cognitive Science, Vol. 14, No. 2, pp. 179-211, 1990. https://doi.org/10.1207/s15516709cog1402_1
  12. M. I. Jordan, Attractor Dynamics and Parallelism in a Connectionist Sequential Machine, Proc. CogSci, pp. 531-546, 1986.
  13. S. Hochreiter and J. Schmidhuber, "Long Short-Term Memory," Neural Comput., Vol. 9, No. 8, pp. 1735- 1780, 1997. https://doi.org/10.1162/neco.1997.9.8.1735
  14. A. Graves, S. Fernandez, F. Gomez, and J. Schmidhuber, Connectionist Temporal Classification: Labelling Unsegmented Sequence Data with Recurrent Neural Networks, Proc. ICML, pp. 369-376, 2006.
  15. A. Graves, A. Mohamed, G. Hinton, Speech Recognition with Deep Recurrent Neural Networks, Proc. ICASSP, pp. 6645-6649, 2013.
  16. K. Cho, B. van Merrienboer, C. Gulcehre, D. Bahdanau, F. Bougares, H. Schwenk, and Y. Bengio, Learning Phrase Representations using RNN Encoder- Decoder for Statistical Machine Translation, Proc. EMNLP, pp. 1724-1734, 2014.
  17. J. Chung, C. Gulcehre, K.H. Cho, and Y. Bengio, "Empirical Evalutation of Gated Recurrent Neural Networks on Sequence Modeling," arXiv:1412.3555, 2014.
  18. R. Jozefowics, W. Zaremba, and I. Sutskever, An Empirical Exploration of Recurrent Neural Network Architectures, Proc. ICML, 37, pp. 2342-2350, 2015.
  19. HP Labs India (2013, Jun. 25), Lipi Toolkit for online Handwriting Recognition (HWR) [Online]. Available: http://lipitk.sourceforge.net/ (downloaded 2017, Jan. 20)
  20. M.-C. Lee and S.-B. Cho, "Accelerometer-Based Gesture Recognition Using Hierarchical Recurrent Neural Network with Bidirectional Long Short- Term Memory," Journal of KIISE: Software and Applications, Vol. 39, No. 12, pp. 1005-1011, 2012. (in Korean)