Journal of Electronic Science and Technology, Volume. 22, Issue 3, 100260(2024)

Global-local combined features to detect pain intensity from facial expression images with attention mechanism

Jiang Wu1...2, Yi Shi2, Shun Yan3, and Hong-Mei Yan12,* |Show fewer author(s)
Author Affiliations
  • 1Yangtze Delta Region Institute (Huzhou), University of Electronic Science and Technology of China, Huzhou, 313001, China
  • 2MOE Key Laboratory for Neuroinformation, School of Life Science and Technology, University of Electronic Science and Technology of China, Chengdu, 611731, China
  • 3College of Engineering, University of California at Santa Barbara, Santa Barbara, 93106, USA
  • show less
    Figures & Tables(10)
    Overall pre-processing pipeline: (a) original image, (b) facial localization, (c) facial alignment, and (d) facial cropping.
    Framework of the proposed GLA-CNN.
    Sample frames and their corresponding PSPI scores at the UNBC-McMaster Shoulder Pain database.
    Number of pictures per PSPI code class at the UNBC-McMaster Shoulder Pain database.
    Confusion matrix based on the proposed GLA-CNN model.
    Samples of attention map from “No Pain” to “Strong Pain” by different models at the UNBC-McMaster Shoulder Pain database.
    • Table 1. Number of image for four pain levels at the UNBC-McMaster Shoulder Pain database2 according to the PSPI score.

      View table
      View in Article

      Table 1. Number of image for four pain levels at the UNBC-McMaster Shoulder Pain database2 according to the PSPI score.

      PSPI scorePain levelNumber of images
      0No pain2928
      1Weak pain2909
      2Mild pain2351
      ≥3Strong pain3102
    • Table 2. Comparison of the average performance of proposed GLA-CNN with VGG-16, separated local and global modules under 5-fold cross validation.

      View table
      View in Article

      Table 2. Comparison of the average performance of proposed GLA-CNN with VGG-16, separated local and global modules under 5-fold cross validation.

      ModelAccuracyF1-scoreRecallPrecision
      Baseline(VGG-16)51.72%36.65%34.55%44.36%
      LANet55.53%35.38%33.78%42.92%
      GANet55.51%37.22%33.91%43.03%
      GLA-CNN56.45%36.52%34.08%43.23%
      GLA-CNN-BiLSTM54.54%33.15%31.07%42.02%
    • Table 3. Comparison of the proposed model with the state-of-the-art at the UNBC-McMaster Shoulder Pain database.

      View table
      View in Article

      Table 3. Comparison of the proposed model with the state-of-the-art at the UNBC-McMaster Shoulder Pain database.

      Ref.MethodLevelAccuracy
      [42]Head analysis428.10%
      [43]LBP434.70%
      [43]LPQ423.40%
      [43]BSIF423.70%
      [44]CNN435.30%
      [16]CNN451.10%
      [45]MRN446.18%
      [46]MA-NET455.16%
      [47]Swin transformer454.40%
      [48]POSTER V2452.24%
      This workGLA-CNN456.45%
    • Table 4. Comparison of different network performance with/without attentional mechanisms (GA and LA stand for global attention and local attention, respectively).

      View table
      View in Article

      Table 4. Comparison of different network performance with/without attentional mechanisms (GA and LA stand for global attention and local attention, respectively).

      ModelGALAAccuracyF1-scoreRecallPrecision
      VGG-16××51.72%36.65%34.55%44.36%
      GNet××53.42%33.06%30.18%41.32%
      GANet×55.51%37.22%33.91%43.03%
      LNet××53.28%35.76%33.07%43.77%
      LANet×55.53%35.38%33.78%42.92%
      GLA-CNN56.45%36.52%34.08%43.23%
    Tools

    Get Citation

    Copy Citation Text

    Jiang Wu, Yi Shi, Shun Yan, Hong-Mei Yan. Global-local combined features to detect pain intensity from facial expression images with attention mechanism[J]. Journal of Electronic Science and Technology, 2024, 22(3): 100260

    Download Citation

    EndNote(RIS)BibTexPlain Text
    Save article for my favorites
    Paper Information

    Category:

    Received: Sep. 2, 2023

    Accepted: May. 21, 2024

    Published Online: Oct. 11, 2024

    The Author Email: Yan Hong-Mei (hmyan@uestc.edu.cn)

    DOI:10.1016/j.jnlest.2024.100260

    Topics