Attention to Warp: Deep Metric Learning for Multivariate Time Series

被引:1
作者
Matsuo, Shinnosuke [1 ]
Wu, Xiaomeng [2 ]
Atarsaikhan, Gantugs [1 ]
Kimura, Akisato [2 ]
Kashino, Kunio [2 ]
Iwana, Brian Kenji [1 ]
Uchida, Seiichi [1 ]
机构
[1] Kyushu Univ, Fukuoka, Japan
[2] NTT Corp, Commun Sci Labs, Tokyo, Japan
来源
DOCUMENT ANALYSIS AND RECOGNITION, ICDAR 2021, PT III | 2021年 / 12823卷
关键词
Attention model; Dynamic time warping; Metric learning; Signature verification; ONLINE SIGNATURE VERIFICATION; ALGORITHM;
D O I
10.1007/978-3-030-86334-0_23
中图分类号
TP18 [人工智能理论];
学科分类号
081104 ; 0812 ; 0835 ; 1405 ;
摘要
Deep time series metric learning is challenging due to the difficult trade-off between temporal invariance to nonlinear distortion and discriminative power in identifying non-matching sequences. This paper proposes a novel neural network-based approach for robust yet discriminative time series classification and verification. This approach adapts a parameterized attention model to time warping for greater and more adaptive temporal invariance. It is robust against not only local but also large global distortions, so that even matching pairs that do not satisfy the monotonicity, continuity, and boundary conditions can still be successfully identified. Learning of this model is further guided by dynamic time warping to impose temporal constraints for stabilized training and higher discriminative power. It can learn to augment the inter-class variation through warping, so that similar but different classes can be effectively distinguished. We experimentally demonstrate the superiority of the proposed approach over previous non-parametric and deep models by combining it with a deep online signature verification framework, after confirming its promising behavior in single-letter handwriting classification on the Unipen dataset.
引用
收藏
页码:350 / 365
页数:16
相关论文
共 40 条
  • [1] Usage of autoencoders and Siamese networks for online handwritten signature verification
    Ahrabian, Kian
    BabaAli, Bagher
    [J]. NEURAL COMPUTING & APPLICATIONS, 2019, 31 (12) : 9321 - 9334
  • [2] [Anonymous], 2018, INT C LEARN REPR
  • [3] The great time series classification bake off: a review and experimental evaluation of recent algorithmic advances
    Bagnall, Anthony
    Lines, Jason
    Bostrom, Aaron
    Large, James
    Keogh, Eamonn
    [J]. DATA MINING AND KNOWLEDGE DISCOVERY, 2017, 31 (03) : 606 - 660
  • [4] Bahdanau D, 2016, Arxiv, DOI arXiv:1409.0473
  • [5] Bromley J., 1993, International Journal of Pattern Recognition and Artificial Intelligence, V7, P669, DOI 10.1142/S0218001493000339
  • [6] Che Z., 2017, WORKSHOP MINING LEAR
  • [7] Chuang Li, 2019, 2019 International Conference on Document Analysis and Recognition (ICDAR). Proceedings, P526, DOI 10.1109/ICDAR.2019.00090
  • [8] Human Motion Analysis with Deep Metric Learning
    Coskun, Huseyin
    Tan, David Joseph
    Conjeti, Sailesh
    Navab, Nassir
    Tombari, Federico
    [J]. COMPUTER VISION - ECCV 2018, PT XIV, 2018, 11218 : 693 - 710
  • [9] A new algorithm for identity verification based on the analysis of a handwritten dynamic signature
    Cpalka, Krzysztof
    Zalasinski, Marcin
    Rutkowski, Leszek
    [J]. APPLIED SOFT COMPUTING, 2016, 43 : 47 - 56
  • [10] On-line signature recognition based on VQ-DTW
    Faundez-Zanuy, Marcos
    [J]. PATTERN RECOGNITION, 2007, 40 (03) : 981 - 992