한국음향학회지 (The Journal of the Acoustical Society of Korea)

  • 제42권4호
  • /
  • Pages.329-344
  • /
  • 2023
  • /
  • 1225-4428(pISSN)
  • /
  • 2287-3775(eISSN)
한국음향학회 (The Acoustical Society of Korea)
권호 표지
DOI QR코드

DOI QR Code

소량 및 불균형 능동소나 데이터세트에 대한 딥러닝 기반 표적식별기의 종합적인 분석

Comprehensive analysis of deep learning-based target classifiers in small and imbalanced active sonar datasets

  • 투고 : 2023.06.09
  • 심사 : 2023.06.24
  • 발행 : 2023.07.31
PDF 다운로드
⟨ 이전 논문 다음 논문 ⟩

초록

본 논문에서는 소량 및 불균형 능동소나 데이터세트에 적용된 다양한 딥러닝 기반 표적식별기의 일반화 성능을 종합적으로 분석하였다. 서로 다른 시간과 해역에서 수집된 능동소나 실험 데이터를 이용하여 두 가지 능동소나 데이터세트를 생성하였다. 데이터세트의 각 샘플은 탐지 처리 이후 탐지된 오디오 신호로부터 추출된 시간-주파수 영역 이미지이다. 표적식별기의 신경망 모델은 다양한 구조를 가지는 22개의 Convolutional Neural Networks(CNN) 모델을 사용하였다. 실험에서 두 가지 데이터세트는 학습/검증 데이터세트와 테스트 데이터세트로 번갈아 가며 사용되었으며, 표적식별기 출력의 변동성을 계산하기 위해 학습/검증/테스트를 10번 반복하고 표적식별 성능을 분석하였다. 이때 학습을 위한 초매개변수는 베이지안 최적화를 이용하여 최적화하였다. 실험 결과 본 논문에서 설계한 얕은 층을 가지는 CNN 모델이 대부분의 깊은 층을 가지는 CNN 모델보다 견실하면서 우수한 일반화 성능을 가지는 것을 확인하였다. 본 논문은 향후 딥러닝 기반 능동소나 표적식별 연구에 대한 방향성을 설정할 때 유용하게 사용될 수 있다.

In this study, we comprehensively analyze the generalization performance of various deep learning-based active sonar target classifiers when applied to small and imbalanced active sonar datasets. To generate the active sonar datasets, we use data from two different oceanic experiments conducted at different times and ocean. Each sample in the active sonar datasets is a time-frequency domain image, which is extracted from audio signal of contact after the detection process. For the comprehensive analysis, we utilize 22 Convolutional Neural Networks (CNN) models. Two datasets are used as train/validation datasets and test datasets, alternatively. To calculate the variance in the output of the target classifiers, the train/validation/test datasets are repeated 10 times. Hyperparameters for training are optimized using Bayesian optimization. The results demonstrate that shallow CNN models show superior robustness and generalization performance compared to most of deep CNN models. The results from this paper can serve as a valuable reference for future research directions in deep learning-based active sonar target classification.

키워드

과제정보

본 연구는 국방과학연구소의 연구비 지원(과제번호:UD210005DD)과 2023년도 정부(교육부)의 재원으로 한국연구재단의 지원(No. 2022R1A6A3A01087548)을 받아 수행됨.

참고문헌

  1. D. A. Abraham, Underwater Acoustic Signal Processing: Modeling, Detection, and Estimation (Springer, Berlin, 2019), Ch. 1.
  2. R. Harrison, C. Yang, C.-F. Lin, T. Politopoulos, and E. Chang, "Classification of underwater targets with active sonar," Proc. IEEE Regional Conf. Aerospace Control Systems, 534-538 (1993).
  3. R. P. Gorman and T. J. Sejnowski, "Analysis of hidden units in a layered network trained to classify sonar targets," Neural networks, 1, 75-89 (1988). https://doi.org/10.1016/0893-6080(88)90023-8
  4. S. M. Murphy and P. C. Hines, "Examining the robustness of automated aural classification of active sonar echoes," J. Acoust. Soc. Am. 135, 626-636 (2014). https://doi.org/10.1121/1.4861922
  5. I. Seo, S. Kim, Y. Ryu, J. Park, and D. S. Han, "Underwater moving target classification using multilayer processing of active sonar system," Appl. Sci. 9, 4617 (2019).
  6. T. Sun, J. Jin, T. Liu, and J. Zhang, "Active sonar target classification method based on fisher's dictionary learning," Appl. Sci. 11, 10635 (2021).
  7. A. Voulodimos, N. Doulamis, A. Doulamis, and E. Protopapadakis, "Deep learning for computer vision: A brief review," Comput. Intell. Neurosci. 2018, 1-13 (2018).
  8. Z. Zhang, J. Geiger, J. Pohjalainen, A. E. De. Mousa, W. Jin, and B. Schuller, "Deep learning for environmentally robust speech recognition: An overview of recent developments," ACM, Trans. Intell. Syst. Tech. 9, 1-28, (2018).
  9. K. Muhammad, A. Ullah, J. Lloret, J. D. Ser, and V. H. C. de Albuquerque, "Deep learning for safe autonomous driving: Current challenges and future directions," IEEE Trans. Intell. Transp. Syst. 22, 4316-4336 (2020).
  10. I. Lauriola, A. Lavelli, and F. Aiolli. "An introduction to deep learning in natural language processing: Models, techniques, and tools," Neurocomputing, 470, 443-456 (2022). https://doi.org/10.1016/j.neucom.2021.05.103
  11. I. Goodfellow, Y. Bengio, and A. Courville, Deep Learning (MIT press, Cambridge, 2016), pp. 1-800.
  12. G. De Magistris, P. Stinco, J. R. Bates, J. M. Topple, G. Canepa, G. Ferri, A. Tesei, and K. Le Page, "Automatic object classification for low-frequency active sonar using convolutional neural networks," Proc. OCEANS, 1-6 (2019).
  13. S. Lee, I. Seo, J. Seok, Y. Kim, and D. S. Han, "Active sonar target classification with power-normalized cepstral coefficients and convolutional neural network," Appl. Sci. 10, 8450 (2020).
  14. Y. Chen, H. Liang, and S. Pang, "Study on small samples active sonar target recognition based on deep learning," J. Mar. Sci. Eng. 10, 1144 (2022).
  15. P. Stinco, G. De Magistris, A. Tesei, and K. D. LePage, "Automatic object classification with active sonar using unsupervised anomaly detection," Proc. EUSIPCO, 46-50 (2021).
  16. G. Kim and Y. Choo, "Bi-sphere anomaly detection with learnable centroid for active sonar classification," IEEE Access, 10, 128590-128603 (2022). https://doi.org/10.1109/ACCESS.2022.3227646
  17. R. O. Nielsen, Sonar Signal Processing (Artech House, Boston, 1991), pp. 1-384.
  18. A. D. Waite, Sonar for Practising Engineers (Wiley, New Jersey, 2002), pp. 1-298.
  19. S. K. Mitra, Digital Signal Processing: A Computer-Based Approach 2nd Ed. (McGraw-Hill, NewYork, 2001), pp. 767-770.
  20. S. G. K. Patro and K. K. Sahu, "Normalization: A preprocessing stage," arXiv preprint arXiv:1503.06462, 1-4 (2015).
  21. L. McInnes, J. Healy, and J. Melville. "Umap: Uniform manifold approximation and pro-jection for dimension reduction," arXiv preprint arXiv:1802. 03426, 1-63 (2018).
  22. Z. Dai, H. Liu, Q. V. LE, and M. Tan, "Coatnet: Marrying convolution and attention for all data sizes," Proc. NeurIPS, 3965-3977 (2021).
  23. S. Marcel and Y. Rodriguez, "Torchvision the machine-vision package of torch," Proc. 18th ACM international conf. Multimedia, 1485-1488 (2010).
  24. O. Russakovsky, J. Deng, H. Su, J. Krause, S. Satheesh, S. Ma, Z. Huang, A. Karpathy, A. Khosla, M. Bernstein, A. C. Berg, and L. Fei-Fei, "Imagenet large scale visual recognition challenge," Proc. IJCV, 211-252 (2015).
  25. A. Geron, Hands-On Machine Learning with Scikit-Learn, Keras, and Tensorflow (O'Reilly Media, California, 2022), pp. 1-856.
  26. D. P. Kingma and J. Ba, "Adam: A method for stochastic optimization," Proc. ICLR, 1-15 (2014).
  27. Bayesian Optimization: Open Source Constrained Global Optimization Tool for Python, https://github.com/bayesian-optimization/BayesianOptimization, (Last viewed May 20, 2023).
  28. Y. Choo, K. Lee, W. Hong, S.-H. Byun, and H. Yang, "Active underwater target detection using a shallow neural network with spectrogram-based temporal variation features," IEEE J. Ocean. Eng. (Early Access) 1-15 (2022).
  29. D. P. Williams, "On the use of tiny convolutional neural networks for human-expert-level classification performance in sonar imagery," IEEE J. Ocean. Eng. 46, 236-260 (2021). https://doi.org/10.1109/JOE.2019.2963041