KIPS Transactions on Software and Data Engineering (정보처리학회논문지:소프트웨어 및 데이터공학)

Korea Information Processing Society (한국정보처리학회)
권호 표지
DOI QR코드

DOI QR Code

Line-Segment Feature Analysis Algorithm for Handwritten-Digits Data Reduction

필기체 숫자 데이터 차원 감소를 위한 선분 특징 분석 알고리즘

  • 김창민 (상지대학교 컴퓨터공학과) ;
  • 이우범 (상지대학교 정보통신소프트웨어공학과)
  • Received : 2020.09.14
  • Accepted : 2020.12.17
  • Published : 2021.04.30
Download PDF
⟨ Previous Next ⟩

Abstract

As the layers of artificial neural network deepens, and the dimension of data used as an input increases, there is a problem of high arithmetic operation requiring a lot of arithmetic operation at a high speed in the learning and recognition of the neural network (NN). Thus, this study proposes a data dimensionality reduction method to reduce the dimension of the input data in the NN. The proposed Line-segment Feature Analysis (LFA) algorithm applies a gradient-based edge detection algorithm using median filters to analyze the line-segment features of the objects existing in an image. Concerning the extracted edge image, the eigenvalues corresponding to eight kinds of line-segment are calculated, using 3×3 or 5×5-sized detection filters consisting of the coefficient values, including [0, 1, 2, 4, 8, 16, 32, 64, and 128]. Two one-dimensional 256-sized data are produced, accumulating the same response values from the eigenvalue calculated with each detection filter, and the two data elements are added up. Two LFA256 data are merged to produce 512-sized LAF512 data. For the performance evaluation of the proposed LFA algorithm to reduce the data dimension for the recognition of handwritten numbers, as a result of a comparative experiment, using the PCA technique and AlexNet model, LFA256 and LFA512 showed a recognition performance respectively of 98.7% and 99%.

인공신경망의 계층의 깊이가 깊어지고 입력으로 사용되는 데이터 차원이 증가됨에 신경망의 학습 및 인식에 있어서 많은 연산을 고속으로 요구하는 고연산의 문제가 발생한다. 따라서 본 논문에서는 신경망 입력 데이터의 차원을 감소시키기 위한 데이터 차원 감소 방법을 제안한다. 제안하는 선분 특징 분석(Line-segment Feature Analysis; LFA) 알고리즘은 한 영상 내에 존재하는 객체의 선분(Line-segment) 특징을 분석하기 위하여 메디안 필터(median filter)를 사용한 기울기 기반의 윤곽선 검출 알고리즘을 적용한다. 추출된 윤곽 영상은 [0, 1, 2, 4, 8, 16, 32, 64, 128]의 계수 값으로 구성된 3×3 또는 5×5 크기의 검출 필터를 이용하여 8가지 선분의 종류에 상응하는 고유값을 계산한다. 각각의 검출필터로 계산된 고유값으로부터 동일한 반응값을 누적하여 두 개의 1차원의 256 크기의 데이터를 생성하고 두 가지 데이터 요소를 합산하여 LFA256 데이터를, 두 데이터를 합병하여 512 크기의 LAF512 데이터를 생성한다. 제안한 LFA 알고리즘의 성능평가는 필기체 숫자 인식을 위한 데이터 차원 감소를 목적으로 PCA 기법과 AlexNet 모델을 이용하여 비교 실험한 결과 LFA256과 LFA512가 각각 98.7%와 99%의 인식 성능을 보였다.

Keywords

References

  1. S. M. Park, K. B. Sim, and H. G. Yeom, "CNN-based classification of brain connectivity using motor imagery," Journal of Korean Institute of Intelligent Systems, Vol.29, No.2, pp.124-129, 2019. https://doi.org/10.5391/jkiis.2019.29.2.124
  2. S. Lek and J. F. Guegan, "Artificial neural networks as a tool in ecological modelling, an introduction," Ecological modelling, Vol.120, No.2-3, pp.65-73, 1999. https://doi.org/10.1016/S0304-3800(99)00092-7
  3. Y. T. Kwak, Y. G. Lee, and O. S. Kwon, "Modified error back propagation algorithm using the approximating of the hidden nodes in multi-layer perceptron," Journal of Korea Information Science Society: Software and Applications, Vol.28, No.9, pp.603-611, 2001.
  4. O. J. Kwon and S. Y. Bang, "Design of fault tolerant multilayer perceptron through selective augmentation," Journal of Korea Information Science Society: Software and Applications, Vol.24, No.7, pp.694-704, 1997.
  5. A. Veit, M. Wilber, and S. Belongie, "Residual networks behave like ensembles of relatively shallow networks," In Advances in Neural Information Processing Systems. pp.550-558, 2016.
  6. A. Veit, M. Wilber, and S. Belongie, "Residual networks behave like ensembles of relatively shallow networks," In Advances in Neural Information Processing Systems. pp.550-558, 2016.
  7. B. S. Hua, M. K. Tran, and S. K. Yeung, "Pointwise convolutional neural networks." In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, pp.984-993, 2018.
  8. R. Zhao, Y. Hu, J. Dotzel, C. D. Sa, and Z. Zhang, "Building efficient deep neural networks with unitary group convolutions," In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, pp.11303-11312, 2019.
  9. Y. Duan, C. Yang, H. Chen, W. Yan, and H. Li, "Low-complexity point cloud denoising for LiDAR by PCA-based dimension reduction," Optics Communications, 126567, 2020. https://doi.org/10.1016/j.optcom.2020.126567
  10. R. Sheikh, M. Patel, and A. Sinhal, "Recognizing MNIST handwritten data set using PCA and LDA," International Conference on Artificial Intelligence: Advances and Applications 2019, Springer, Singapore, pp.169-177, 2020.
  11. J. Ma and Y. Yuan, "Dimension reduction of image deep feature using PCA," Journal of Visual Communication and Image Representation, Vol.63, pp.102578, 2019. https://doi.org/10.1016/j.jvcir.2019.102578
  12. L. Ali, C. Zhu, Z. Zhang, and Y. Liu, "Automated detection of Parkinson's disease based on multiple types of sustained phonations using linear discriminant analysis and genetically optimized neural network," IEEE Journal of Translational Engineering in Health and Medicine, Vol.7, pp.1-10, 2019.
  13. M. N. I. Qureshi, J. Oh, and B. Lee, "3D-CNN based discrimination of schizophrenia using resting-state fMRI," Artificial Intelligence in Medicine, Vol.98, pp.10-17, 2019. https://doi.org/10.1016/j.artmed.2019.06.003
  14. Y. Dong, and S. J. Qin, "A novel dynamic PCA algorithm for dynamic data modeling and process monitoring," Journal of Process Control, Vol.67, pp.1-11, 2018. https://doi.org/10.1016/j.jprocont.2017.05.002
  15. G. Cohen, S. Afshar, J. Tapson, and A. van Schaik, "EMNIST: an extension of MNIST to handwritten letters." arXiv preprint arXiv:1702.05373, 2017.
  16. Y. Huang, X. Qiao, P. Ren, L. Liu, C. Pu, and J. Chen, "A lightweight collaborative recognition system with binary convolutional neural network for mobile Web augmented reality." In 2019 IEEE 39th International Conference on Distributed Computing Systems, pp.1497-1506, 2019.