pith. sign in

arxiv: 2606.31349 · v1 · pith:AND2ABTOnew · submitted 2026-06-30 · 📡 eess.SP · cs.AI

PGUDA: Pressure-Guided Unsupervised Domain Adaptation with Cross-Modal Knowledge Distillation for sEMG-Based Gesture Recognition

Yurui Liu , Xiao-Cong Zhong , Qisong Wang , Xuefu Wang , Dan Liu , Jinwei Sun This is my paper

Pith reviewed 2026-07-01 04:26 UTC · model grok-4.3

classification 📡 eess.SP cs.AI
keywords sEMGgesture recognitiondomain adaptationknowledge distillationpressure signalscross-subject classificationcross-session classificationlabel efficiency
0
0 comments X

The pith

Pressure signals train a teacher network that distills stable semantics to an sEMG student network for unsupervised adaptation across subjects and sessions.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper introduces PGUDA to overcome performance drops in sEMG gesture recognition caused by distribution shifts across subjects and recording sessions. Conventional domain adaptation struggles with the stochastic nature of sEMG signals and the lack of target labels. PGUDA instead trains a teacher on pressure signals, which carry more consistent physical semantics, and uses cross-modal knowledge distillation to guide an sEMG student network on unlabeled target data. This regularizes representation learning toward modality-invariant features. On a dataset from eleven subjects the method reaches 58.08 percent average accuracy in cross-subject and cross-session tasks, exceeds prior DA baselines, and matches fully supervised performance when the teacher sees only 5 percent labeled examples.

Core claim

PGUDA trains a teacher network on pressure signals and applies cross-modal knowledge distillation so the teacher guides an sEMG student network on unlabeled target domains, thereby transferring consistent physical semantics that regularize representation learning and produce leading accuracies of 58.08 percent in both cross-subject and cross-session gesture classification while requiring only 5 percent labeled data for the teacher.

What carries the argument

Cross-modal knowledge distillation in which a pressure-trained teacher network supplies modality-invariant physical semantics to regularize an sEMG student network on unlabeled target domains.

If this is right

  • PGUDA substantially outperforms existing domain adaptation approaches on both cross-subject and cross-session tasks.
  • Classification accuracy comparable to fully supervised benchmarks is reached with only 5 percent labeled data used to train the teacher network.
  • The framework reduces the amount of labeled calibration data required for practical sEMG gesture recognition systems.
  • It mitigates the effects of sEMG stochasticity by importing stable semantics from the pressure modality.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • The same teacher-student pressure-to-sEMG transfer could be tested on other noisy biosignals such as EEG if a paired stable modality exists.
  • Collecting short paired pressure-sEMG recordings at setup time may enable broader deployment of unsupervised adaptation without ongoing labels.
  • The observed label efficiency suggests the approach could scale to larger user populations with lower annotation cost than fully supervised alternatives.

Load-bearing premise

Pressure signals supply robust, stable physical semantics that can be transferred via knowledge distillation to regularize sEMG representation learning on unlabeled target domains.

What would settle it

Running the same cross-subject and cross-session experiments with the pressure teacher removed or replaced by an sEMG-only teacher and observing no accuracy gain or a drop relative to standard unsupervised DA baselines would falsify the claimed benefit of the cross-modal guidance.

Figures

Figures reproduced from arXiv: 2606.31349 by Dan Liu, Jinwei Sun, Qisong Wang, Xiao-Cong Zhong, Xuefu Wang, Yurui Liu.

Figure 1
Figure 1. Figure 1: t-SNE visualization of feature distributions for different physiological [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Overview of the proposed PGUDA framework. The left panel presents the multimodal data acquisition and preprocessing pipeline. The middle [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: The architecture of the sEMG student network. [PITH_FULL_IMAGE:figures/full_fig_p005_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Data acquisition details. (a) The hardware architecture system; (b) [PITH_FULL_IMAGE:figures/full_fig_p006_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: t-SNE visualization of feature distributions with different methods. The top row is the task of Subject 2 [PITH_FULL_IMAGE:figures/full_fig_p008_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: Ablation study results of the PGUDA. Subplots (a) and (b) illustrate the classification Accuracy and [PITH_FULL_IMAGE:figures/full_fig_p009_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: Error rate comparison (Accuracy and F1-score) under different labeled data ratios. The proposed PGUDA (red) consistently achieves lower error rates than the Supervised baseline (blue) across all settings, demonstrating high label efficiency and robust generalization. the teacher model, while the sEMG student network is trained via cross-modal distillation from synchronized pressure-sEMG pairs without expli… view at source ↗
Figure 8
Figure 8. Figure 8: Impact of the trade-off hyperparameter α on model performance. The dual-axis plot displays Accuracy and F1-score for both cross-subject and cross-session protocols. the baseline achieves 20.00% with 80% labels, a marginal difference of only 0.55%. This demonstrates the high label efficiency of PGUDA. Furthermore, under an extremely low label ratio (1%), PGUDA still significantly reduces the error rate (fro… view at source ↗
read the original abstract

Surface electromyography (sEMG)-based gesture recognition has emerged as a promising technology for natural human-computer interaction. However, its practical deployment remains challenging due to severe performance degradation caused by feature distribution discrepancies across different subjects and recording sessions. Although domain adaptation (DA) techniques are commonly employed to mitigate such discrepancies, conventional methods often struggle to effectively aligning sEMG features, primarily due to their inherent stochasticity and the scarcity of labeled data. To address these limitations, this paper proposes a novel Pressure-Guided Unsupervised Domain Adaptation (PGUDA) framework, which leverages the robustness and stability of pressure signals to introduce a cross-modal knowledge distillation strategy that transfers consistent physical semantics across modalities. Specifically, a teacher network trained on pressure signals guides an sEMG student network on unlabeled target domains, thereby regularizing the representation learning process with transferable and modality-invariant knowledge. Extensive experiments conducted on a self-collected multimodal dataset involving eleven subjects validate the effectiveness of the proposed PGUDA framework. The results demonstrate that our proposed PGUDA achieves leading performance in both cross-subject and cross-session classification tasks, achieving average accuracies of 58.08% and substantially outperforming existing DA approaches. Notably, PGUDA exhibits remarkable label efficiency: it attains classification accuracy comparable to fully supervised benchmarks while requiring only 5% of labeled data for teacher network training. This framework offers a robust and data-efficient solution that can significantly reduce the calibration burden in practical sEMG-based gesture recognition systems.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

3 major / 2 minor

Summary. The paper proposes PGUDA, a framework for sEMG-based gesture recognition that performs unsupervised domain adaptation by training a teacher network on pressure signals and using cross-modal knowledge distillation to regularize an sEMG student network on unlabeled target domains. It reports leading results on a self-collected multimodal dataset of 11 subjects, with 58.08% average accuracy in cross-subject and cross-session tasks that substantially exceeds existing DA baselines, plus label efficiency where comparable accuracy is reached using only 5% labeled data for the teacher.

Significance. If the central assumption holds and the empirical results are reproducible, the work would offer a practical route to lower calibration effort in sEMG systems by exploiting an auxiliary pressure modality for regularization. The reported label-efficiency result would be a notable strength if substantiated with proper controls.

major comments (3)
  1. [Abstract] Abstract: the claim that pressure signals supply 'robustness and stability' that can be transferred via knowledge distillation rests on an untested assumption; no variance comparison, ablation, or isolated experiment demonstrates that pressure exhibits lower cross-subject or cross-session shift than sEMG, which is load-bearing for the teacher-student regularization argument.
  2. [Abstract] Abstract / Experiments section: all quantitative claims (58.08% accuracy, outperformance of DA baselines, and 5%-label equivalence to fully supervised performance) are presented without error bars, number of runs, statistical tests, or description of baseline implementations and hyper-parameter selection, rendering the 'leading performance' and 'remarkable label efficiency' assertions unverifiable.
  3. [Abstract] Abstract: the self-collected dataset of eleven subjects is the sole empirical basis for every result, yet the collection protocol (sensor placement, gesture set, session structure, subject demographics) is not described, preventing assessment of whether the reported domain shifts are representative or whether the pressure modality's purported stability is an artifact of the acquisition setup.
minor comments (2)
  1. [Method] Notation for the teacher-student loss and distillation temperature is introduced without an explicit equation or algorithmic listing, making the precise form of the cross-modal regularization difficult to reproduce.
  2. [Abstract] The abstract states 'extensive experiments' but provides no table or figure reference for the per-subject or per-session breakdowns that would support the averaged 58.08% figure.

Simulated Author's Rebuttal

3 responses · 0 unresolved

We thank the referee for the constructive feedback, which highlights important areas for improving the clarity and rigor of our manuscript. We address each major comment below and commit to revisions that directly respond to the concerns raised.

read point-by-point responses

  1. Referee: [Abstract] Abstract: the claim that pressure signals supply 'robustness and stability' that can be transferred via knowledge distillation rests on an untested assumption; no variance comparison, ablation, or isolated experiment demonstrates that pressure exhibits lower cross-subject or cross-session shift than sEMG, which is load-bearing for the teacher-student regularization argument.

    Authors: We acknowledge that the current manuscript does not include a direct empirical comparison of cross-subject or cross-session variance between pressure and sEMG modalities, which would better substantiate the core assumption. While the motivation is grounded in prior work on pressure sensing being less prone to certain physiological variations, we agree this requires explicit validation. In the revised version, we will add an ablation study that computes and compares feature variances across subjects and sessions for both modalities on our dataset, along with a quantitative analysis of how pressure guidance affects the student's representation stability. revision: yes

  2. Referee: [Abstract] Abstract / Experiments section: all quantitative claims (58.08% accuracy, outperformance of DA baselines, and 5%-label equivalence to fully supervised performance) are presented without error bars, number of runs, statistical tests, or description of baseline implementations and hyper-parameter selection, rendering the 'leading performance' and 'remarkable label efficiency' assertions unverifiable.

    Authors: We agree that the results lack the necessary statistical details and implementation transparency to allow full verification. In the revision, we will report all accuracies as means with standard deviations computed over multiple independent runs (minimum of five random seeds), include error bars in all relevant figures, conduct and report statistical significance tests (such as paired t-tests or Wilcoxon tests) against the baselines, and add a subsection detailing the re-implementation of each baseline, the hyperparameter search procedure, and the validation-based selection process. revision: yes

  3. Referee: [Abstract] Abstract: the self-collected dataset of eleven subjects is the sole empirical basis for every result, yet the collection protocol (sensor placement, gesture set, session structure, subject demographics) is not described, preventing assessment of whether the reported domain shifts are representative or whether the pressure modality's purported stability is an artifact of the acquisition setup.

    Authors: The full collection protocol is described in Section 4.1 of the manuscript (sensor placement on the forearm, the set of 10 gestures, multi-session recording structure per subject, and subject demographics). However, we recognize that the abstract provides insufficient context and that additional specifics (e.g., exact sensor models and sampling rates) could be clarified. We will therefore expand the abstract with a brief summary of the dataset and augment Section 4.1 with any omitted acquisition parameters to ensure the domain shifts and modality properties can be properly evaluated. revision: yes

Circularity Check

0 steps flagged

No circularity: empirical results on held-out data with no self-referential derivations

full rationale

The paper proposes an empirical PGUDA framework and reports accuracies (e.g., 58.08% average) from experiments on a self-collected multimodal dataset with eleven subjects, using cross-subject and cross-session splits. No equations, fitted parameters, or self-citations are presented that reduce the claimed performance or label-efficiency results to inputs by construction. The derivation chain consists of standard knowledge-distillation training followed by evaluation on held-out domains; the outcomes remain falsifiable against external benchmarks and do not collapse into tautological redefinitions or self-citation load-bearing steps.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 0 invented entities

The framework rests on the domain assumption that pressure signals are sufficiently stable to serve as a teacher for sEMG; no free parameters or invented entities are explicitly introduced in the abstract.

axioms (1)
  • domain assumption Pressure signals are robust and stable relative to sEMG and therefore supply transferable physical semantics for cross-modal distillation.
    Invoked to justify training the teacher on pressure and distilling to the sEMG student on unlabeled targets.

pith-pipeline@v0.9.1-grok · 5816 in / 1284 out tokens · 40040 ms · 2026-07-01T04:26:47.921059+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

47 extracted references · 40 canonical work pages · 5 internal anchors

  1. [1]

    A review of hand function rehabilitation systems based on hand motion recognition devices and artificial intelligence,

    Y . Gu, Y . Xu, Y . Shen, H. Huang, T. Liu, L. Jin, H. Ren, and J. Wang, “A review of hand function rehabilitation systems based on hand motion recognition devices and artificial intelligence,”Brain Sciences, vol. 12, no. 8, p. 1079, Aug. 2022. [Online]. Available: http://dx.doi.org/10.3390/brainsci12081079

    work page doi:10.3390/brainsci12081079 2022

  2. [2]

    A survey on hand gesture recognition based on surface electromyography: Fundamentals, methods, applications, challenges and future trends,

    S. Ni, M. A. Al-qaness, A. Hawbani, D. Al-Alimi, M. Abd Elaziz, and A. A. Ewees, “A survey on hand gesture recognition based on surface electromyography: Fundamentals, methods, applications, challenges and future trends,”Applied Soft Computing, vol. 166, p. 112235, Nov

  3. [3]

    Available: http://dx.doi.org/10.1016/j.asoc.2024.112235

    [Online]. Available: http://dx.doi.org/10.1016/j.asoc.2024.112235

    work page doi:10.1016/j.asoc.2024.112235 2024

  4. [4]

    A multi-stream convolutional neural network for semg-based gesture recognition in muscle-computer interface,

    W. Wei, Y . Wong, Y . Du, Y . Hu, M. Kankanhalli, and W. Geng, “A multi-stream convolutional neural network for semg-based gesture recognition in muscle-computer interface,”Pattern Recognition Letters, vol. 119, p. 131–138, Mar. 2019. [Online]. Available: http://dx.doi.org/ 10.1016/j.patrec.2017.12.005

    work page doi:10.1016/j.patrec.2017.12.005 2019

  5. [5]

    A robust, real-time control scheme for multifunction myoelectric control,

    K. Englehart and B. Hudgins, “A robust, real-time control scheme for multifunction myoelectric control,”IEEE Transactions on Biomedical Engineering, vol. 50, no. 7, p. 848–854, Jul. 2003. [Online]. Available: http://dx.doi.org/10.1109/TBME.2003.813539

    work page doi:10.1109/tbme.2003.813539 2003

  6. [6]

    Support vector machine-based classification scheme for myoelectric control applied to upper limb,

    M. Oskoei and H. Hu, “Support vector machine-based classification scheme for myoelectric control applied to upper limb,”IEEE Transactions on Biomedical Engineering, vol. 55, no. 8, p. 1956–1965, Aug. 2008. [Online]. Available: http://dx.doi.org/10.1109/TBME.2008. 919734

    work page doi:10.1109/tbme.2008 1956

  7. [7]

    Empowering hand rehabilitation with ai-powered gesture recognition: A study of an semg-based system,

    K. Guo, M. Orban, J. Lu, M. S. Al-Quraishi, H. Yang, and M. Elsamanty, “Empowering hand rehabilitation with ai-powered gesture recognition: A study of an semg-based system,”Bioengineering, vol. 10, no. 5, p. 557, May 2023. [Online]. Available: http: //dx.doi.org/10.3390/bioengineering10050557

    work page doi:10.3390/bioengineering10050557 2023

  8. [8]

    Stcn-gr: Spatial-temporal convolutional networks for surface-electromyography-based gesture recognition,

    Z. Lai, X. Kang, H. Wang, W. Zhang, X. Zhang, P. Gong, L. Niu, and H. Huang, “Stcn-gr: Spatial-temporal convolutional networks for surface-electromyography-based gesture recognition,” inNeural Infor- mation Processing. Springer International Publishing, 2021, p. 27–39. [Online]. Available: http://dx.doi.org/10.1007/978-3-030-92238-2 3

    work page doi:10.1007/978-3-030-92238-2 2021

  9. [9]

    Bioformers: Embedding transformers for ultra-low power semg-based gesture recognition,

    A. Burrello, F. B. Morghet, M. Scherer, S. Benatti, L. Benini, E. Macii, M. Poncino, and D. J. Pagliari, “Bioformers: Embedding transformers for ultra-low power semg-based gesture recognition,” in2022 Design, Automation; Test in Europe Conference; Exhibition (DATE). IEEE, Mar. 2022, p. 1443–1448. [Online]. Available: http://dx.doi.org/10.23919/DATE54114.2...

    work page doi:10.23919/date54114.2022.9774639 2022

  10. [10]

    Distilling the Knowledge in a Neural Network

    G. Hinton, O. Vinyals, and J. Dean, “Distilling the knowledge in a neural network,” 2015. [Online]. Available: https://arxiv.org/abs/1503.02531

    work page internal anchor Pith review Pith/arXiv arXiv 2015

  11. [11]

    Knowledge distillation: A survey,

    J. Gou, B. Yu, S. J. Maybank, and D. Tao, “Knowledge distillation: A survey,”International Journal of Computer Vision, vol. 129, no. 6, p. 1789–1819, Mar. 2021. [Online]. Available: http://dx.doi.org/10.1007/ s11263-021-01453-z

    2021

  12. [12]

    Cross modal distillation for supervision transfer,

    S. Gupta, J. Hoffman, and J. Malik, “Cross modal distillation for supervision transfer,” in2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2016, pp. 2827–2836

    2016

  13. [13]

    Cmcrd: Cross- modal contrastive representation distillation for emotion recognition,

    S. Kan, H. Wu, Z. Cui, F. Huang, X. Xu, and D. Wu, “Cmcrd: Cross- modal contrastive representation distillation for emotion recognition,”

  14. [14]

    Available: https://arxiv.org/abs/2504.09221

    [Online]. Available: https://arxiv.org/abs/2504.09221

    work page arXiv

  15. [15]

    A cross-modal densely guided knowledge distillation based on modality rebalancing strategy for enhanced unimodal emotion recognition,

    S. Wu, H. Liang, Y . Zhang, Y . Chen, and Z. Jia, “A cross-modal densely guided knowledge distillation based on modality rebalancing strategy for enhanced unimodal emotion recognition,” inProceedings of the Thirty-Fourth International Joint Conference on Artificial Intelligence, ser. IJCAI-2025. International Joint Conferences on Artificial Intelligence O...

    work page doi:10.24963/ijcai.2025/472 2025

  16. [16]

    Emotionkd: A cross-modal knowledge distillation framework for emotion recognition based on physiological signals,

    Y . Liu, Z. Jia, and H. Wang, “Emotionkd: A cross-modal knowledge distillation framework for emotion recognition based on physiological signals,” inProceedings of the 31st ACM International Conference on Multimedia, ser. MM ’23. ACM, Oct. 2023, p. 6122–6131. [Online]. Available: http://dx.doi.org/10.1145/3581783.3612277

    work page doi:10.1145/3581783.3612277 2023

  17. [17]

    Efficient hand gesture recognition using multi-task multi-modal learning and self-distillation,

    J.-Y . Li, H. Prawiro, C.-C. Chiang, H.-Y . Chang, T.-Y . Pan, C.-T. Huang, and M.-C. Hu, “Efficient hand gesture recognition using multi-task multi-modal learning and self-distillation,” inACM Multimedia Asia 2023, ser. MMAsia ’23. ACM, Dec. 2023, p. 1–7. [Online]. Available: http://dx.doi.org/10.1145/3595916.3626411

    work page doi:10.1145/3595916.3626411 2023

  18. [18]

    Cross-modal knowledge distillation for continuous sign language recognition,

    L. Gao, P. Shi, L. Hu, J. Feng, L. Zhu, L. Wan, and W. Feng, “Cross-modal knowledge distillation for continuous sign language recognition,”Neural Networks, vol. 179, p. 106587, Nov. 2024. [Online]. Available: http://dx.doi.org/10.1016/j.neunet.2024.106587

    work page doi:10.1016/j.neunet.2024.106587 2024

  19. [19]

    A deep domain adaptation framework with correlation alignment for eeg-based motor imagery classification,

    X.-C. Zhong, Q. Wang, D. Liu, J.-X. Liao, R. Yang, S. Duan, G. Ding, and J. Sun, “A deep domain adaptation framework with correlation alignment for eeg-based motor imagery classification,”Computers in Biology and Medicine, vol. 163, p. 107235, Sep. 2023. [Online]. Available: http://dx.doi.org/10.1016/j.compbiomed.2023.107235

    work page doi:10.1016/j.compbiomed.2023.107235 2023

  20. [20]

    Cross modality knowledge distillation between a-mode ultrasound and surface electromyography,

    J. Zeng, Y . Sheng, Y . Yang, Z. Zhou, and H. Liu, “Cross modality knowledge distillation between a-mode ultrasound and surface electromyography,”IEEE Transactions on Instrumentation and Measurement, vol. 71, p. 1–9, 2022. [Online]. Available: http://dx.doi.org/10.1109/TIM.2022.3195264

    work page doi:10.1109/tim.2022.3195264 2022

  21. [21]

    A transferable adaptive domain adversarial neural network for virtual reality augmented emg-based gesture recognition,

    U. Cote-Allard, G. Gagnon-Turcotte, A. Phinyomark, K. Glette, E. Scheme, F. Laviolette, and B. Gosselin, “A transferable adaptive domain adversarial neural network for virtual reality augmented emg-based gesture recognition,”IEEE Transactions on Neural Systems and Rehabilitation Engineering, vol. 29, p. 546–555, 2021. [Online]. Available: http://dx.doi.or...

    work page doi:10.1109/tnsre.2021.3059741 2021

  22. [22]

    Cross- modal learning for domain adaptation in 3d semantic segmentation,

    M. Jaritz, T.-H. Vu, R. de Charette, E. Wirbel, and P. Perez, “Cross- modal learning for domain adaptation in 3d semantic segmentation,” IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. 45, no. 2, p. 1533–1544, Feb. 2023. [Online]. Available: http://dx.doi.org/10.1109/TPAMI.2022.3159589

    work page doi:10.1109/tpami.2022.3159589 2023

  23. [23]

    Imu2doppler: Cross-modal domain adaptation for doppler-based activity recognition using imu data,

    S. Bhalla, M. Goel, and R. Khurana, “Imu2doppler: Cross-modal domain adaptation for doppler-based activity recognition using imu data,”Proc. ACM Interact. Mob. Wearable Ubiquitous Technol., vol. 5, no. 4, Dec. 2022. [Online]. Available: https://doi.org/10.1145/3494994

    work page doi:10.1145/3494994 2022

  24. [24]

    Optimization of inter-subject semg-based hand gesture recognition tasks using unsupervised domain adaptation techniques,

    Z. Wang, H. Wan, L. Meng, Z. Zeng, M. Akay, C. Chen, and W. Chen, “Optimization of inter-subject semg-based hand gesture recognition tasks using unsupervised domain adaptation techniques,”Biomedical Signal Processing and Control, vol. 92, p. 106086, Jun. 2024. [Online]. Available: http://dx.doi.org/10.1016/j.bspc.2024.106086

    work page doi:10.1016/j.bspc.2024.106086 2024

  25. [25]

    A multi-scale cnn for transfer learning in semg-based hand gesture recognition for prosthetic devices,

    R. Fratti, N. Marini, M. Atzori, H. M ¨uller, C. Tiengo, and F. Bassetto, “A multi-scale cnn for transfer learning in semg-based hand gesture recognition for prosthetic devices,”Sensors, vol. 24, no. 22, p. 7147, Nov. 2024. [Online]. Available: http://dx.doi.org/10.3390/s24227147

    work page doi:10.3390/s24227147 2024

  26. [26]

    Multisource adversarial feature disentanglement method for cross- subject gesture recognition using semg signals,

    K. Su, K. Liu, B. Wan, H. Qiao, J. Huang, M. Feng, and J. Liu, “Multisource adversarial feature disentanglement method for cross- subject gesture recognition using semg signals,”IEEE Transactions on Instrumentation and Measurement, vol. 74, p. 1–12, 2025. [Online]. Available: http://dx.doi.org/10.1109/TIM.2025.3557823

    work page doi:10.1109/tim.2025.3557823 2025

  27. [27]

    High-density emg grip force estimation during muscle fatigue via domain adaptation,

    H. Pan, D. Li, C. Chen, and P. B. Shull, “High-density emg grip force estimation during muscle fatigue via domain adaptation,” IEEE Transactions on Neural Systems and Rehabilitation Engineering, vol. 33, p. 925–934, 2025. [Online]. Available: http://dx.doi.org/10. 1109/TNSRE.2025.3541227

    work page arXiv 2025

  28. [28]

    Deep end-to-end transfer learning for robust inter-subject and inter-day hand gesture recognition using surface emg,

    J. Li, X. Jiang, J. Fan, Y . Geng, F. Jia, and C. Dai, “Deep end-to-end transfer learning for robust inter-subject and inter-day hand gesture recognition using surface emg,”Biomedical Signal Processing and Control, vol. 100, p. 106892, Feb. 2025. [Online]. Available: http://dx.doi.org/10.1016/j.bspc.2024.106892

    work page doi:10.1016/j.bspc.2024.106892 2025

  29. [29]

    Deep learning for emg-based human-machine interaction: A review,

    D. Xiong, D. Zhang, X. Zhao, and Y . Zhao, “Deep learning for emg-based human-machine interaction: A review,”IEEE/CAA Journal of Automatica Sinica, vol. 8, no. 3, p. 512–533, Mar. 2021. [Online]. Available: http://dx.doi.org/10.1109/JAS.2021.1003865

    work page doi:10.1109/jas.2021.1003865 2021

  30. [30]

    Exploration of feature extraction methods and dimension for semg signal classification,

    Y . Wu, X. Hu, Z. Wang, J. Wen, J. Kan, and W. Li, “Exploration of feature extraction methods and dimension for semg signal classification,”Applied Sciences, vol. 9, no. 24, p. 5343, Dec. 2019. [Online]. Available: http://dx.doi.org/10.3390/app9245343

    work page doi:10.3390/app9245343 2019

  31. [31]

    Framework for learning a hand intent recognition model from semg for fes-based control,

    N. Das, S. Endo, H. Kavianirad, and S. Hirche, “Framework for learning a hand intent recognition model from semg for fes-based control,” in2024 10th IEEE RAS/EMBS International Conference for Biomedical Robotics and Biomechatronics (BioRob). IEEE, Sep. 2024, p. 1320–1327. [Online]. Available: http://dx.doi.org/10.1109/ BioRob60516.2024.10719910 12

    work page arXiv 2024

  32. [32]

    Deep learning for electromyographic hand gesture signal classification using transfer learning,

    U. Cote-Allard, C. L. Fall, A. Drouin, A. Campeau-Lecours, C. Gosselin, K. Glette, F. Laviolette, and B. Gosselin, “Deep learning for electromyographic hand gesture signal classification using transfer learning,”IEEE Transactions on Neural Systems and Rehabilitation Engineering, vol. 27, no. 4, p. 760–771, Apr. 2019. [Online]. Available: http://dx.doi.org...

    work page doi:10.1109/tnsre.2019.2896269 2019

  33. [33]

    A Simple Framework for Contrastive Learning of Visual Representations

    T. Chen, S. Kornblith, M. Norouzi, and G. Hinton, “A simple framework for contrastive learning of visual representations,” 2020. [Online]. Available: https://arxiv.org/abs/2002.05709

    work page internal anchor Pith review Pith/arXiv arXiv 2020

  34. [34]

    Adaptive batch normalization for practical domain adaptation,

    Y . Li, N. Wang, J. Shi, X. Hou, and J. Liu, “Adaptive batch normalization for practical domain adaptation,”Pattern Recognition, vol. 80, p. 109–117, Aug. 2018. [Online]. Available: http://dx.doi.org/ 10.1016/j.patcog.2018.03.005

    work page doi:10.1016/j.patcog.2018.03.005 2018

  35. [35]

    A kernel two-sample test,

    A. Gretton, K. M. Borgwardt, M. J. Rasch, B. Sch ¨olkopf, and A. Smola, “A kernel two-sample test,”J. Mach. Learn. Res., vol. 13, no. 1, p. 723–773, Mar. 2012

    2012

  36. [36]

    B. Sun, J. Feng, and K. Saenko,Correlation Alignment for Unsupervised Domain Adaptation. Springer International Publishing, 2017, p. 153–171. [Online]. Available: http://dx.doi.org/10.1007/ 978-3-319-58347-1 8

    2017

  37. [37]

    Domain-Adversarial Training of Neural Networks

    Y . Ganin, E. Ustinova, H. Ajakan, P. Germain, H. Larochelle, F. Laviolette, M. Marchand, and V . Lempitsky, “Domain-adversarial training of neural networks,” 2015. [Online]. Available: https: //arxiv.org/abs/1505.07818

    work page internal anchor Pith review Pith/arXiv arXiv 2015

  38. [38]

    Deep subdomain adaptation network for image classification,

    Y . Zhu, F. Zhuang, J. Wang, G. Ke, J. Chen, J. Bian, H. Xiong, and Q. He, “Deep subdomain adaptation network for image classification,” IEEE Transactions on Neural Networks and Learning Systems, vol. 32, no. 4, p. 1713–1722, Apr. 2021. [Online]. Available: http://dx.doi.org/10.1109/tnnls.2020.2988928

    work page doi:10.1109/tnnls.2020.2988928 2021

  39. [39]

    Towards discriminability and diversity: Batch nuclear-norm maximization under label insufficient situations,

    S. Cui, S. Wang, J. Zhuo, L. Li, Q. Huang, and Q. Tian, “Towards discriminability and diversity: Batch nuclear-norm maximization under label insufficient situations,” in2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR). IEEE, Jun. 2020, p. 3940–3949. [Online]. Available: http://dx.doi.org/10.1109/ CVPR42600.2020.00400

    work page arXiv 2020

  40. [40]

    Maximum classifier discrepancy for unsupervised domain adaptation,

    K. Saito, K. Watanabe, Y . Ushiku, and T. Harada, “Maximum classifier discrepancy for unsupervised domain adaptation,” in2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition. IEEE, Jun. 2018, p. 3723–3732. [Online]. Available: http://dx.doi.org/10.1109/ CVPR.2018.00392

    work page arXiv 2018

  41. [41]

    Conditional Adversarial Domain Adaptation

    M. Long, Z. Cao, J. Wang, and M. I. Jordan, “Conditional adversarial domain adaptation,” 2017. [Online]. Available: https: //arxiv.org/abs/1705.10667

    work page internal anchor Pith review Pith/arXiv arXiv 2017

  42. [42]

    Bridging Theory and Algorithm for Domain Adaptation

    Y . Zhang, T. Liu, M. Long, and M. I. Jordan, “Bridging theory and algorithm for domain adaptation,” 2019. [Online]. Available: https://arxiv.org/abs/1904.05801

    work page internal anchor Pith review Pith/arXiv arXiv 2019

  43. [43]

    Aligning domain-specific distribution and classifier for cross-domain classification from multiple sources,

    Y . Zhu, F. Zhuang, and D. Wang, “Aligning domain-specific distribution and classifier for cross-domain classification from multiple sources,” Proceedings of the AAAI Conference on Artificial Intelligence, vol. 33, no. 01, p. 5989–5996, Jul. 2019. [Online]. Available: http://dx.doi.org/10.1609/aaai.v33i01.33015989

    work page doi:10.1609/aaai.v33i01.33015989 2019

  44. [44]

    A comprehensive survey on source-free domain adaptation,

    J. Li, Z. Yu, Z. Du, L. Zhu, and H. T. Shen, “A comprehensive survey on source-free domain adaptation,”IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. 46, no. 8, p. 5743–5762, Aug. 2024. [Online]. Available: http://dx.doi.org/10.1109/tpami.2024.3370978

    work page doi:10.1109/tpami.2024.3370978 2024

  45. [45]

    Plug-and-play semg-driven hand gesture recognition with subdomain adaptation for exoskeleton rehabilitation gloves,

    X.-C. Zhong, Q. Wang, D. Liu, X. Wang, R. Li, Y . Wang, M. Zhang, and J. Sun, “Plug-and-play semg-driven hand gesture recognition with subdomain adaptation for exoskeleton rehabilitation gloves,”IEEE Transactions on Instrumentation and Measurement, vol. 74, p. 1–10,

  46. [46]

    Available: http://dx.doi.org/10.1109/TIM.2024.3502881

    [Online]. Available: http://dx.doi.org/10.1109/TIM.2024.3502881

    work page doi:10.1109/tim.2024.3502881 2024

  47. [47]

    A review of domain adaptation without target labels,

    W. M. Kouw and M. Loog, “A review of domain adaptation without target labels,”IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. 43, no. 3, p. 766–785, Mar. 2021. [Online]. Available: http://dx.doi.org/10.1109/TPAMI.2019.2945942

    work page doi:10.1109/tpami.2019.2945942 2021