Pith. sign in

REVIEW 3 major objections 6 minor 98 references

A physics-based self-supervised decoder trains IMU sensing with zero labels and beats supervised models.

Reviewed by Pith at T0; open to challenge. T0 means a machine referee read the full paper against a public rubric. the ladder, T0–T4 →

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

T0 review · deepseek-v4-flash

2026-08-04 04:09 UTC pith:G7YOI25S

load-bearing objection Label-free IMU sensing with a genuinely new architecture and strong benchmark numbers, but the disentanglement of sensor motion from body motion is not proven, and the results have no error bars; still deserves a serious referee. the 3 major comments →

arxiv 2607.18361 v2 pith:G7YOI25S submitted 2026-07-20 cs.LG cs.AI

Physical Self-Supervised Learning: IMU Sensing without Manual Labels

Yuyang Leng , Renyuan Liu , Shaohan Hu , Peijun Zhao , Chun-Fu Chen (Richard) , Songqing Chen , Shuochao Yao This is my paper
classification cs.LG cs.AI
keywords IMU sensingself-supervised learninginertial trackingmotion capturephysics-informed decoderkinematic treesensor placementlabel-free learning
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved

The pith

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

This paper argues that IMU-based inertial tracking and full-body motion capture can be trained with no manual labels at all, replacing the conventional neural decoder with an auto-adaptive physics decoder — learnable kinematic equations that reconstruct IMU readings from predicted body motion. The reason this matters is that labeling IMU data today requires lab instrumentation or extra sensor modalities, and models degrade when devices move or placements change; a label-free approach that encodes known physics could make sensing models cheap to deploy across new devices, users, and wearing conditions. The authors report that their framework reduces tracking error by up to 5x and motion-capture error by up to 4x in challenging generalization settings, and that it consistently outperforms supervised and self-supervised baselines without any labels. The central bet is that physical structure, not labeled data, is the missing supervision signal.

Core claim

The central claim is that a self-supervised autoencoder becomes a complete IMU sensing system when its decoder is a learnable family of kinematic equations rather than a black-box network. The encoder predicts physical states (joint rotations, global translation and orientation, bone lengths) and an environment-aware representation; the physics decoder turns those states back into IMU readings, and training minimizes reconstruction error in a denoised latent space. To make reconstruction unambiguous, the framework separates sensor motion from body motion using probabilistic frequency-spatial constraints — body motion is band-limited (a 25 Hz cutoff captures more than 99% of the energy) while

What carries the argument

The load-bearing component is the auto-adaptive physics decoder, a differentiable forward model of IMU kinematics: given predicted object states and learned environment variables (bone lengths, sensor placement, and sensor-relative motion), it reconstructs accelerometer and gyroscope readings through discrete-time kinematic equations. Around it, the probabilistic frequency-spatial constraints force sensor-relative motion to stay within bounded ranges and body motion to be band-limited, which is what makes the sensor-versus-body disentanglement tractable; the multi-view kinematic tree then lets sparse IMU anchors supervise every joint; and the uncertainty-aware distributional formulation prop

Load-bearing premise

The method assumes human body motion is band-limited below about 25 Hz and that sensor-relative motion stays within the hand-set ranges of Table 1; if either fails, reconstruction can be satisfied by moving the sensor instead of the body, and the predicted physical states are no longer trustworthy.

What would settle it

Take an IMU device through motions that violate the spatial bounds — a phone thrown loosely in a bag or a watch spinning freely on the wrist — while recording ground-truth body motion with an external system; if reconstruction loss stays low while pose or trajectory error climbs, the frequency-spatial constraints are not enforcing disentanglement, and the central claim fails.

Watch this falsifier. Get emailed when new claim-graph text bears on it.

Share X Bluesky LinkedIn Reddit HN

If this is right

  • No labeled data means IMU models can be trained or retrained for a new device, placement, or user from raw recordings alone, eliminating the 10–20% labeled effort that prior self-supervised methods still need for domain adaptation.
  • The method's advantage grows exactly where sensors are not rigidly attached: under loose wearing, supervised baselines collapse while the framework keeps over 90% of poses within 25 degrees of error.
  • Sparse setups with as few as a phone, watch, or earbud remain usable, and the reported gap over baselines widens as sensors get sparser.
  • A lightweight variant runs on a smartphone and an embedded microcontroller within real-time budgets while keeping the lowest reported errors.
  • Reconstructed skeletons transfer to downstream tasks such as activity recognition, gait recognition, and fall detection, serving as a generic motion representation.

Where Pith is reading between the lines

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

  • If the claim holds, the standard pretrain-then-finetune recipe for IMU sensing may be unnecessary; a plausible next step, not explored here, is continuous on-device adaptation from unlabeled streams using the same self-supervised objective.
  • The hand-set spatial bounds in Table 1 rest on empirical experience; the paper itself leaves deriving tighter, data-driven bounds to future work, which would be a natural way to test how much of the gain depends on these priors.
  • The frequency-spatial disentanglement creates a sharp, testable boundary: devices that violate the bounds (a phone tumbling in a bag, a watch spinning on the wrist) should cause the model to explain away body motion as sensor motion, and measuring where performance collapses would map the method's valid operating envelope.
  • If the result generalizes, label-free skeleton estimation could become a generic representation layer for mobile sensing, letting downstream tasks inherit robustness without learning it from scarce labels.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit.

Referee Report

3 major / 6 minor

Summary. The paper proposes a self-supervised autoencoder framework for IMU sensing (inertial tracking and full-body motion capture) that requires no manual labels. The decoder is replaced by an auto-adaptive physics decoder built from discretized kinematic equations, with parameters predicted from a two-stage hybrid IMU encoder; training uses reconstruction in a denoised latent space. Additional components include probabilistic frequency–spatial constraints for disentangling sensor and object motion, a multi-view kinematic tree for propagating sparse supervision, and an uncertainty-aware Monte Carlo formulation. Experiments are reported on TotalCapture, DIP-IMU, Nymeria, SHL, OxIOD, and a self-collected dataset, with claims of up to 5× tracking and 4× motion-capture error reductions over supervised baselines.

Significance. If the claims hold, the paper would be a significant contribution to mobile sensing: it offers a label-free training paradigm for two practically important IMU tasks and aims to improve robustness to sensor placement and looseness. The paper is well structured, ships code, evaluates on multiple public benchmarks, and includes a sensitivity analysis of the frequency and spatial priors. These are genuine strengths. However, the empirical evidence is weakened by the absence of uncertainty quantification and by an unanalyzed identifiability issue in the core disentanglement mechanism.

major comments (3)
  1. [§5, Tables 2–3] All reported results are single point estimates; no standard deviations, confidence intervals, or number of seeds/trials are given. Many headline comparisons are small in absolute terms: e.g., in Table 2 on TotalCapture, Ours-Lite reports a SIP error of 12.69 versus Ours at 12.66, and in Table 4 the angular error changes only from 13.08 to 13.12 across spatial scales. These differences are likely within run-to-run noise. To support the claim of 'consistently outperforming', the authors should provide repeated-run statistics and, where possible, paired significance tests.
  2. [§3.2, Table 1, Eq. (8)] The disentanglement of sensor-relative motion from object motion is not shown to be identifiable. For the 'Backpack' placement, Table 1 marks rotation as 'Unlimited' and bounds translation at ±10 cm. Because §3.2 states that sensor-relative motion is intentionally not band-limited, a static body with the phone rotating/translating inside the bag can produce the same IMU sequence as a moving body with a fixed sensor; the reconstruction loss cannot distinguish these explanations. The sensitivity analysis in Table 4 sweeps α and λ only; it does not test whether the model silently assigns motion to the sensor instead of the body. The paper should either give a formal identifiability argument under the stated constraints or provide a targeted experiment with gold-standard sensor-relative motion (e.g., synthetic data or an external tracker on the phone), especially for the 'Unlimited' rotation
  3. [§5.2 and Abstract] The headline 'up to 5×/4×' improvements are not backed by tabulated numbers. Figures 14–17 present leave-one-condition-out results only as bar charts, without numeric values, baseline numbers, or confidence intervals. The abstract's quantitative claims should be tied to reproducible numbers in tables or a supplementary file. Additionally, the self-collected dataset has only four participants, and the paper does not specify how leave-one-scenario-out splits are constructed (subject vs. session), making the generalization claims difficult to evaluate.
minor comments (6)
  1. [Table 3] The header 'Oxiod' should read 'OxIOD' for consistency with the text.
  2. [Figure 2] The figure is missing axis labels and units. The text states that a 25 Hz cutoff captures >99% of energy, but the per-dataset percentages are not reported, making the claim hard to verify.
  3. [§5.4, Figure 20] The text below the figure contains garbled fragments such as 'HHar AMAS S OurDataset'; this should be cleaned up.
  4. [Table 1] The entry 'Earbud Designated ear' should be phrased clearly, e.g., 'designated ear' as the only placement candidate.
  5. [§4.1] Only four participants were recruited for the self-collected data. The paper should report more detail on participant variability and the number of trials per condition.
  6. [§5.6] For a 6-second window, Ours-Lite takes 3.462 s on the ARM Cortex-M7 for MoCap, which is not strictly real-time. The text should qualify what 'real-time' means for this platform.

Circularity Check

0 steps flagged

No significant circularity; claims are empirical benchmark results, not derivations that reduce to their inputs.

full rationale

The paper's central claim is an empirical performance comparison: a self-supervised autoencoder with an auto-adaptive physics decoder and frequency–spatial constraints is trained without labels and evaluated on held-out ground truth from public datasets (DIP-IMU, TotalCapture, Nymeria, SHL, OxIOD) and its own Kinect-synchronized collection. No predicted headline number is a fitted constant renamed as a prediction. The spatial bounds in Table 1 are explicitly hand-set priors ('chosen based on empirical experience rather than tuned for optimality'), and Table 4 sweeps both the cutoff frequency and spatial scale over a plateau, showing the reported results are not forced by a fitted hyperparameter. The 25 Hz band-limit is justified by external biomechanics citations and an empirical energy analysis (Figure 2), not by the target labels. The physics decoder is structured as kinematic equations (Eqs. 1–8) rather than a trainable black box, but that is architectural prior knowledge, not a circular definition: the reconstruction loss is an unsupervised training signal, and the test-time outputs are compared against labels never used in training. The paper honestly acknowledges that IMU inference is underdetermined ('A single IMU time series can correspond to multiple, distinct human motion trajectories') and that the sensor-placement bounds are experiential; those are correctness/generalizability risks (identifiability of the sensor/body decomposition is not proven), not circularity. Self-citations appear only in related-work context (e.g., DeepSense, SenseGAN, DeepIoT) and are not load-bearing for the framework or its claims. Because the evaluation is externally benchmarked and the constraints are not fitted to the evaluation labels, the derivation chain is self-contained and no circular step can be exhibited.

Axiom & Free-Parameter Ledger

3 free parameters · 5 axioms · 0 invented entities

The paper introduces no new physical entities (no new particles, forces, or conserved quantities). Its contributions are architectural and procedural: the auto-adaptive physics decoder, multi-view kinematic tree, and frequency-spatial priors. These are testable via benchmarks, so independent evidence comes from the reported evaluations, but the hand-set priors and sampling defaults are the main free parameters.

free parameters (3)
  • Spatial constraint bounds (α, λ) per sensor placement = e.g., watch ±30°, ±1cm/±3cm; phone ±40°/±3cm; earbud fixed; backpack ±10cm (Table 1)
    Hand-set based on 'empirical experience' (Section 3.2); used to bound the tanh activation in Eq. (8). Sensitivity analysis (Table 4) shows robustness, so they act as priors rather than fragile fits; however, they are free parameters chosen by the authors, not measured per deployment.
  • Frequency cutoff (25 Hz default) = 25 Hz (outputs at 50 Hz Nyquist)
    Chosen to capture >99% of human motion energy (Section 3.2, Figure 2). Sensitivity 20-40 Hz gives small variation (Table 4), so it is a structural regularizer rather than a tightly tuned constant.
  • Monte Carlo sample count for uncertainty propagation = not stated; Figure 19 tests 2-16 samples
    Sampling density trades computation against reliability (Section 5.3.2). The main results do not state the default sample count, making exact replication harder.
axioms (5)
  • domain assumption Human body motion is band-limited: a 25 Hz cutoff retains >99% of motion energy
    Section 3.2 and Figure 2; empirically verified on four public mocap datasets, but not guaranteed for all motions (e.g., impacts, tremors). Load-bearing for the frequency constraint.
  • ad hoc to paper Sensor-relative motion is spatially bounded by the ranges in Table 1 (watch ±30°, phone ±40°, etc.)
    Section 3.2/Table 1; chosen from 'empirical experience', not measured. If a sensor moves beyond these bounds (e.g., phone thrown in a bag, watch spinning), the disentanglement breaks and the reconstruction may be explained by sensor motion alone.
  • domain assumption SMPL forward kinematics is an adequate differentiable model of the human body for these tasks
    Section 3.1.2; standard in the field, but limits accuracy for non-SMPL bodies or non-human objects and assumes the kinematic tree structure.
  • domain assumption Reconstruction in the frozen IMUProj latent space preserves task-relevant information
    Section 3.1.3; the paper asserts noise reduction from MAE pretraining, but does not prove the latent retains all information needed for pose/tracking; this is an architectural bet.
  • standard math Gumbel-Softmax reparameterization gives a faithful differentiable approximation to discrete placement sampling
    Sections 3.2 and 3.4; standard technique in deep learning, used for placement selection and uncertainty-aware sampling.

pith-pipeline@v1.3.0-alltime-deepseek · 23202 in / 10291 out tokens · 330531 ms · 2026-08-04T04:09:17.693797+00:00 · methodology

0 comments
Cite this review

Pith. "Pith review of Physical Self-Supervised Learning: IMU Sensing without Manual Labels." pith.science (2026). https://pith.science/paper/G7YOI25S

@misc{pith2026260718361,
  author       = {Pith},
  title        = {Pith review of: Physical Self-Supervised Learning: IMU Sensing without Manual Labels},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/G7YOI25S}},
  note         = {Machine review of arXiv:2607.18361}
}
Share X Bluesky LinkedIn Reddit HN
read the original abstract

Deep neural networks have become a promising approach for IMU-based sensing, but their scalability is fundamentally limited by costly labeled data and poor robustness to heterogeneous devices, placements, and users. Existing unsupervised and self-supervised methods reduce but do not remove this dependence, still requiring labeled data for domain adaptation and largely ignoring known physical structure. We propose physical self-supervised learning, an autoencoder-style paradigm for label-free IMU sensing. We replace the conventional neural decoder with an auto-adaptive physics decoder, a learnable family of kinematic equations that enforces explicit physical structure while adapting across environments, and adopt a hybrid two-stage IMU encoder with reconstruction in a structured latent space to mitigate sensor noise. Our framework further introduces probabilistic frequency-spatial constraints to disentangle sensor and object motion, a multi-view kinematic tree to exploit sparse physical self-supervised signals, and an uncertainty-aware formulation to handle the inherent ambiguity of IMU inference. Evaluated on inertial tracking and full-body motion capture over public datasets and realistic deployments, physical self-supervised learning reduces errors by up to 5x for tracking and 4x for motion capture in challenging generalization scenarios, consistently outperforming state-of-the-art supervised and self-supervised baselines without any labels. Our code is available at https://github.com/YuyangLeng/physical-ssl-imu-label-free

Figures

Figures reproduced from arXiv: 2607.18361 by Chun-Fu Chen (Richard), Peijun Zhao, Renyuan Liu, Shaohan Hu, Shuochao Yao, Songqing Chen, Yuyang Leng.

Figure 1
Figure 1. Figure 1: Physical Self-Supervised Learning. In practice, however, many IMU-sensing tasks involve objects with substantially more complex kinematic structure, such as hu￾man arms or full-body motion, where each IMU measurement re￾flects the coupled dynamics of an articulated rigid-body system rather than a single point mass. We model such an articulated object as a kinematic tree [68]: a tree-structured assembly of … view at source ↗
Figure 3
Figure 3. Figure 3: Probabilistic frequency–spatial constraints for mo [PITH_FULL_IMAGE:figures/full_fig_p005_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Multi-view kinematic tree. Three “detour” kine [PITH_FULL_IMAGE:figures/full_fig_p006_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Definition of wearing tightness. On a cross-sectional [PITH_FULL_IMAGE:figures/full_fig_p007_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: Angular error across different tightness conditions. All other baselines assume ground-truth sensor placements and [PITH_FULL_IMAGE:figures/full_fig_p008_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: Angular error on datasets with 6 sensors. [PITH_FULL_IMAGE:figures/full_fig_p008_7.png] view at source ↗
Figure 8
Figure 8. Figure 8: Angular error distributions shown for tight, normal, [PITH_FULL_IMAGE:figures/full_fig_p008_8.png] view at source ↗
Figure 10
Figure 10. Figure 10: Inertial tracking trajectory estimation. Baselines diverge within minutes due to accumulated error. [PITH_FULL_IMAGE:figures/full_fig_p009_10.png] view at source ↗
Figure 12
Figure 12. Figure 12: Inertial Tracking comparison on public datasets [PITH_FULL_IMAGE:figures/full_fig_p009_12.png] view at source ↗
Figure 14
Figure 14. Figure 14: Leave-one-condition-out evaluation by wearing tightness. Non-shaded bars show out-of-domain performance when [PITH_FULL_IMAGE:figures/full_fig_p010_14.png] view at source ↗
Figure 15
Figure 15. Figure 15: Motion capture results under additional leave-one-scenario-out evaluations, covering both motion and user scenarios. [PITH_FULL_IMAGE:figures/full_fig_p010_15.png] view at source ↗
Figure 16
Figure 16. Figure 16: Inertial tracking results under additional leave-one-scenario-out evaluations, covering both motion and user scenarios. [PITH_FULL_IMAGE:figures/full_fig_p010_16.png] view at source ↗
Figure 17
Figure 17. Figure 17: Leave-One-Dataset-Out Training to Evaluate [PITH_FULL_IMAGE:figures/full_fig_p010_17.png] view at source ↗
Figure 18
Figure 18. Figure 18: Stepwise ablation of our model under different sensor-attachment conditions. Starting from the full model (Ours(+All)), [PITH_FULL_IMAGE:figures/full_fig_p011_18.png] view at source ↗
Figure 19
Figure 19. Figure 19: Output uncertainty vs. predictive error for different sample counts. All cases remain close to the ideal, indicating [PITH_FULL_IMAGE:figures/full_fig_p011_19.png] view at source ↗
Figure 20
Figure 20. Figure 20: Performance on three downstream tasks: HAR, gait recognition, and fall detection. These results show that the [PITH_FULL_IMAGE:figures/full_fig_p012_20.png] view at source ↗

Review history (2 revisions) →

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

98 extracted references · 7 linked inside Pith

  1. [1]

    Sussex-Huawei Locomotion and Transportation Dataset

    2018. Sussex-Huawei Locomotion and Transportation Dataset. doi:10.21227/7vtt- 8c19

    doi:10.21227/7vtt- 2018

  2. [2]

    Hamad Ahmed and Muhammad Tahir. 2016. Accurate attitude estimation of a moving land vehicle using low-cost MEMS IMU sensors.IEEE Transactions on Intelligent Transportation Systems18, 7 (2016), 1723–1739

    2016

  3. [3]

    Karan Ahuja, Andy Kong, Mayank Goel, and Chris Harrison. 2020. Direction- of-voice (dov) estimation for intuitive speech interaction with smart devices ecosystems. InProceedings of the 33rd Annual ACM Symposium on User Interface Software and Technology. 1121–1131

    2020

  4. [4]

    Eric R Bachmann, Robert B McGhee, Xiaoping Yun, and Michael J Zyda. 2001. Inertial and magnetic posture tracking for inserting humans into networked virtual environments. InProceedings of the ACM symposium on Virtual reality software and technology. 9–16

    2001

  5. [5]

    Billur Barshan and Hugh F Durrant-Whyte. 1995. Inertial navigation systems for mobile robots.IEEE transactions on robotics and automation11, 3 (1995), 328–342

    1995

  6. [6]

    Konstantinos Benidis, Syama Sundar Rangapuram, Valentin Flunkert, Yuyang Wang, Danielle Maddix, Caner Turkmen, Jan Gasthaus, Michael Bohlke-Schneider, David Salinas, Lorenzo Stella, et al. 2022. Deep Learning for Time Series Fore- casting: Tutorial and Literature Survey.Comput. Surveys55, 6 (2022), 1–36

    2022

  7. [7]

    Brenda Bigland-Ritchie. 1981. EMG/force relations and fatigue of human volun- tary contractions.Exercise and Sport Sciences Reviews9, 1 (1981), 75–118

    1981

  8. [8]

    Shengze Cai, Zhiping Mao, Zhicheng Wang, Minglang Yin, and George Em Karniadakis. 2021. Physics-informed neural networks (PINNs) for fluid mechanics: A review.Acta Mechanica Sinica37, 12 (2021), 1727–1738

    2021

  9. [9]

    Shengze Cai, Zhicheng Wang, Sifan Wang, Paris Perdikaris, and George Em Karniadakis. 2021. Physics-informed neural networks for heat transfer problems. Journal of Heat Transfer143, 6 (2021), 060801

    2021

  10. [10]

    Andy Catruna, Adrian Cosma, and Emilian Radoi. 2024. GaitPT: Skeletons are All You Need for Gait Recognition. In18th IEEE International Conference on Automatic Face and Gesture Recognition (FG 2024). IEEE, 1–10

    2024

  11. [11]

    Changhao Chen, Xiaoxuan Lu, Andrew Markham, and Niki Trigoni. 2018. Ionet: Learning to cure the curse of drift in inertial odometry. InProceedings of the AAAI Conference on Artificial Intelligence, Vol. 32

    2018

  12. [12]

    Changhao Chen, Peijun Zhao, Chris Xiaoxuan Lu, Wei Wang, Andrew Markham, and Niki Trigoni. 2018. OxIOD: The Dataset for Deep Inertial Odometry. arXiv:1809.07491 [cs.RO] https://arxiv.org/abs/1809.07491

    Pith/arXiv arXiv 2018

  13. [13]

    Taco Cohen and Max Welling. 2016. Group equivariant convolutional networks. InInternational conference on machine learning. PMLR, 2990–2999

    2016

  14. [14]

    Taco S Cohen, Mario Geiger, Jonas Köhler, and Max Welling. 2018. Spherical cnns.arXiv preprint arXiv:1801.10130(2018)

    Pith/arXiv arXiv 2018

  15. [15]

    Xiaoran Fan, Longfei Shangguan, Siddharth Rupavatharam, Yanyong Zhang, Jie Xiong, Yunfei Ma, and Richard Howard. 2021. HeadFi: bringing intelligence to all headphones. InProceedings of the 27th Annual International Conference on Mobile Computing and Networking. 147–159

    2021

  16. [16]

    Geissinger and Alan T

    Jack H. Geissinger and Alan T. Asbeck. 2020. Motion Inference Using Sparse Inertial Sensors, Self-Supervised Learning, and a New Dataset of Unscripted Human Motion.Sensors20, 21 (2020). doi:10.3390/s20216330

    doi:10.3390/s20216330 2020

  17. [17]

    2016.Deep Learning

    Ian Goodfellow, Yoshua Bengio, and Aaron Courville. 2016.Deep Learning. MIT Press. http://www.deeplearningbook.org

    2016

  18. [18]

    Mahanth Gowda, Justin Manweiler, Ashutosh Dhekne, Romit Roy Choudhury, and Justin D Weisz. 2016. Tracking drone orientation with multiple GPS receivers. InProceedings of the 22nd annual international conference on mobile computing and networking. 280–293

    2016

  19. [19]

    Tian Hao, Guoliang Xing, and Gang Zhou. 2013. iSleep: unobtrusive sleep quality monitoring using smartphones. InProceedings of the 11th ACM Conference on Embedded Networked Sensor Systems. 1–14

    2013

  20. [20]

    Harish Haresamudram, Apoorva Beedu, Varun Agrawal, Patrick L Grady, Irfan Essa, Judy Hoffman, and Thomas Plötz. 2020. Masked reconstruction based self-supervision for human activity recognition. InProceedings of the 2020 ACM International Symposium on Wearable Computers. 45–49

    2020

  21. [21]

    Harish Haresamudram, Irfan Essa, and Thomas Plötz. 2021. Contrastive predictive coding for human activity recognition.Proceedings of the ACM on Interactive, Mobile, Wearable and Ubiquitous Technologies5, 2 (2021), 1–26

    2021

  22. [22]

    Kaiming He, Xinlei Chen, Saining Xie, Yanghao Li, Piotr Dollár, and Ross Girshick

  23. [23]

    Yuze He, Chen Bian, Jingfei Xia, Shuyao Shi, Zhenyu Yan, Qun Song, and Guoliang Xing. 2023. Vi-map: Infrastructure-assisted real-time hd mapping for autonomous driving. InProceedings of the 29th Annual International Conference on Mobile Computing and Networking. 1–15

    2023

  24. [24]

    Sachini Herath, Hang Yan, and Yasutaka Furukawa. 2020. Ronin: Robust neural inertial navigation in the wild: Benchmark, evaluations, & new methods. In2020 IEEE international conference on robotics and automation (ICRA). IEEE, 3146–3152

    2020

  25. [25]

    Black, Otmar Hilliges, and Gerard Pons-Moll

    Yinghao Huang, Manuel Kaufmann, Emre Aksan, Michael J. Black, Otmar Hilliges, and Gerard Pons-Moll. 2018. Deep inertial poser: learning to reconstruct human pose from sparse inertial measurements in real time.ACM Trans. Graph.37, 6, Article 185 (Dec. 2018), 15 pages. doi:10.1145/3272127.3275108

    arXiv 2018

  26. [26]

    Eric Jang, Shixiang Gu, and Ben Poole. 2016. Categorical reparameterization with gumbel-softmax.arXiv preprint arXiv:1611.01144(2016)

    Pith/arXiv arXiv 2016

  27. [27]

    Wenjun Jiang, Chenglin Miao, Fenglong Ma, Shuochao Yao, Yaqing Wang, Ye Yuan, Hongfei Xue, Chen Song, Xin Ma, Dimitrios Koutsonikolas, et al . 2018. Towards environment independent device free human activity recognition. In Proceedings of the 24th annual international conference on mobile computing and networking. 289–304

    2018

  28. [28]

    Antonio R Jimenez, Fernando Seco, Carlos Prieto, and Jorge Guevara. 2009. A comparison of pedestrian dead-reckoning algorithms using a low-cost MEMS IMU. In2009 IEEE International Symposium on Intelligent Signal Processing. IEEE, 37–42

    2009

  29. [29]

    Denizhan Kara, Tomoyoshi Kimura, Shengzhong Liu, Jinyang Li, Dongxin Liu, Tianshi Wang, Ruijie Wang, Yizhuo Chen, Yigong Hu, and Tarek Abdelzaher

  30. [30]

    Prerna Khanna, IV Ramakrishnan, Shubham Jain, Xiaojun Bi, and Aruna Bala- subramanian. 2024. Hand Gesture Recognition for Blind Users by Tracking 3D Gesture Trajectory. InProceedings of the CHI Conference on Human Factors in Computing Systems. 1–15

    2024

  31. [31]

    Rushil Khurana, Karan Ahuja, Zac Yu, Jennifer Mankoff, Chris Harrison, and Mayank Goel. 2018. GymCam: Detecting, recognizing and tracking simultaneous exercises in unconstrained scenes.Proceedings of the ACM on Interactive, Mobile, Wearable and Ubiquitous Technologies2, 4 (2018), 1–17

    2018

  32. [32]

    A. I. King. 1984. A Technical Survey: A Review of Biomechanical Models.Journal of Biomechanical Engineering106, 2 (1984), 97–104. doi:10.1115/1.3138480

    doi:10.1115/1.3138480 1984

  33. [33]

    Diederik P Kingma and Max Welling. 2013. Auto-encoding variational bayes. arXiv preprint arXiv:1312.6114(2013)

    Pith/arXiv arXiv 2013

  34. [34]

    Aditi Krishnapriyan, Amir Gholami, Shandian Zhe, Robert Kirby, and Michael W Mahoney. 2021. Characterizing possible failure modes in physics-informed neural networks.Advances in neural information processing systems34 (2021), 26548– 26560

    2021

  35. [35]

    Hyeokhyen Kwon, Catherine Tong, Harish Haresamudram, Yan Gao, Gregory D Abowd, Nicholas D Lane, and Thomas Ploetz. 2020. Imutube: Automatic extrac- tion of virtual on-body accelerometry from video for human activity recognition. Proceedings of the ACM on Interactive, Mobile, Wearable and Ubiquitous Technolo- gies4, 3 (2020), 1–29

    2020

  36. [36]

    Nicholas D Lane and Petko Georgiev. 2015. Can deep learning revolutionize mobile sensing?. InProceedings of the 16th international workshop on mobile computing systems and applications. 117–122

    2015

  37. [37]

    Nicholas D Lane, Petko Georgiev, and Lorena Qendro. 2015. Deepear: robust smartphone audio sensing in unconstrained acoustic environments using deep learning. InProceedings of the 2015 ACM international joint conference on pervasive and ubiquitous computing. 283–294

    2015

  38. [38]

    Hyung-Jik Lee and Seul Jung. 2009. Gyro sensor drift compensation by Kalman fil- ter to control a mobile inverted pendulum robot system. In2009 IEEE International Conference on Industrial Technology. IEEE, 1–6

    2009

  39. [39]

    Yuyang Leng, Renyuan Liu, Hongpeng Guo, Songqing Chen, and Shuochao Yao

  40. [40]

    Zikang Leng, Hyeokhyen Kwon, and Thomas Plötz. 2023. Generating virtual on-body accelerometer data from virtual textual descriptions for human activity recognition. InProceedings of the 2023 ACM International Symposium on Wearable Computers. 39–43

    2023

  41. [41]

    Yuxuan Liang, Haomin Wen, Yuqi Nie, Yushan Jiang, Ming Jin, Dongjin Song, Shirui Pan, and Qingsong Wen. 2024. Foundation Models for Time Series Analysis: A Tutorial and Survey. InProceedings of the 30th ACM SIGKDD Conference on Knowledge Discovery and Data Mining. 6555–6565. doi:10.1145/3637528.3671451

    arXiv 2024

  42. [42]

    Renyuan Liu, Yuyang Leng, Kaiyan Liu, Shaohan Hu, Chun-Fu Chen, Peijun Zhao, Heechul Yun, and Shuochao Yao. 2025. DAF: An Efficient End-to-End Dynamic Activation Framework for on-Device DNN Training. InProceedings of the 23rd Annual International Conference on Mobile Systems, Applications and Services. 196–208

    2025

  43. [43]

    Renyuan Liu, Yuyang Leng, Shilei Tian, Shaohan Hu, Chun-Fu Chen, and Shuochao Yao. 2024. Dynaspa: Exploiting spatial sparsity for efficient dynamic dnn inference on devices. InProceedings of the 22nd ACM Conference on Embedded Networked Sensor Systems. 422–435

    2024

  44. [44]

    Renyuan Liu, Yuyang Leng, Shilei Tian, Shaohan Hu, Richard Chen, and Shuochao Yao. 2025. On-Device Dynamic DNN Inference through Spatial Sparsity Exploita- tion.GetMobile: Mobile Computing and Communications29, 3 (2025), 35–38

    2025

  45. [45]

    Wenxin Liu, David Caruso, Eddy Ilg, Jing Dong, Anastasios I Mourikis, Kostas Daniilidis, Vijay Kumar, and Jakob Engel. 2020. Tlio: Tight learned inertial odometry.IEEE Robotics and Automation Letters5, 4 (2020), 5653–5660

    2020

  46. [46]

    Yang Liu, Zhenjiang Li, Zhidan Liu, and Kaishun Wu. 2019. Real-time arm skeleton tracking and gesture inference tolerant to missing wearable sensors. InProceedings of the 17th Annual International Conference on Mobile Systems, Applications, and Services. 287–299. MobiSys ’26, June 21–25, 2026, Cambridge, United Kingdom Y. Leng, R. Liu, S. Hu, P. Zhao, C.-...

    2019

  47. [47]

    Matthew Loper, Naureen Mahmood, Javier Romero, Gerard Pons-Moll, and Michael J Black. 2023. SMPL: A skinned multi-person linear model. InSeminal Graphics Papers: Pushing the Boundaries, Volume 2. 851–866

    2023

  48. [48]

    Wenjie Luo, Zhenyu Yan, Qun Song, and Rui Tan. 2021. PhyAug: Physics-directed data augmentation for deep sensing model transfer in cyber-physical systems. InProceedings of the 20th International Conference on Information Processing in Sensor Networks (co-located with CPS-IoT Week 2021). 31–46

    2021

  49. [49]

    João Luís Marins, Xiaoping Yun, Eric R Bachmann, Robert B McGhee, and Michael J Zyda. 2001. An extended Kalman filter for quaternion-based orientation estimation using MARG sensors. InProceedings 2001 IEEE/RSJ International Con- ference on Intelligent Robots and Systems. Expanding the Societal Role of Robotics in the the Next Millennium (Cat. No. 01CH3718...

    2001

  50. [50]

    Hossein Mousavi Hondori and Maryam Khademi. 2014. A review on technical and clinical impact of microsoft kinect on physical therapy and rehabilitation. Journal of medical engineering2014, 1 (2014), 846514

    2014

  51. [51]

    Jens Bo Nielsen. 2016. Human spinal motor control.Annual Review of Neuroscience 39, 1 (2016), 81–101

    2016

  52. [52]

    Nadhira Noor and In Kyu Park. 2023. A Lightweight Skeleton-Based 3D-CNN for Real-Time Fall Detection and Action Recognition. InProceedings of the IEEE/CVF International Conference on Computer Vision Workshops. 2179–2188

    2023

  53. [53]

    Xiaomin Ouyang, Xian Shuai, Jiayu Zhou, Ivy Wang Shi, Zhiyuan Xie, Guoliang Xing, and Jianwei Huang. 2022. Cosmo: contrastive fusion learning with small data for multimodal human activity recognition. InProceedings of the 28th Annual International Conference on Mobile Computing And Networking. 324–337

    2022

  54. [54]

    Xiaomin Ouyang, Zhiyuan Xie, Jiayu Zhou, Jianwei Huang, and Guoliang Xing

  55. [55]

    Jose Luis Ponton, Haoran Yun, Andreas Aristidou, Carlos Andujar, and Nuria Pelechano. 2023. SparsePoser: Real-time full-body motion reconstruction from sparse data.ACM Transactions on Graphics43, 1 (2023), 1–14

    2023

  56. [56]

    Hangwei Qian, Tian Tian, and Chunyan Miao. 2022. What makes good contrastive learning on small-scale wearable-based tasks?. InProceedings of the 28th ACM SIGKDD conference on knowledge discovery and data mining. 3761–3771

    2022

  57. [57]

    Setareh Rahimi Taghanaki, Michael J Rainbow, and Ali Etemad. 2021. Self- supervised human activity recognition by learning to predict cross-dimensional motion. InProceedings of the 2021 ACM International Symposium on Wearable Computers. 23–27

    2021

  58. [58]

    Bingbing Rao, Ehsan Kazemi, Yifan Ding, Devu M Shila, Frank M Tucker, and Liqiang Wang. 2022. Ctin: Robust contextual transformer network for inertial navigation. InProceedings of the AAAI Conference on Artificial Intelligence, Vol. 36. 5413–5421

    2022

  59. [59]

    Angelo M Sabatini. 2006. Quaternion-based extended Kalman filter for determin- ing orientation by inertial and magnetic sensing.IEEE transactions on Biomedical Engineering53, 7 (2006), 1346–1356

    2006

  60. [60]

    Vıctor Garcia Satorras, Emiel Hoogeboom, and Max Welling. 2021. E (n) equi- variant graph neural networks. InInternational conference on machine learning. PMLR, 9323–9332

    2021

  61. [61]

    Sheng Shen, Mahanth Gowda, and Romit Roy Choudhury. 2018. Closing the gaps in inertial motion tracking. InProceedings of the 24th Annual International Conference on Mobile Computing and Networking. 429–444

    2018

  62. [62]

    Sheng Shen, He Wang, and Romit Roy Choudhury. 2016. I am a smartwatch and i can track my user’s arm. InProceedings of the 14th annual international conference on Mobile systems, applications, and services. 85–96

    2016

  63. [63]

    Allan Stisen, Henrik Blunck, Sourav Bhattacharya, Thor Siiger Prentow, Mikkel Baun Kjærgaard, Anind Dey, Tobias Sonne, and Mads Møller Jensen

  64. [64]

    Jochen Tautges, Arno Zinke, Björn Krüger, Jan Baumann, Andreas Weber, Thomas Helten, Meinard Müller, Hans-Peter Seidel, and Bernd Eberhardt. 2011. Motion reconstruction using sparse accelerometer data.ACM Transactions on Graphics (ToG)30, 3 (2011), 1–12

    2011

  65. [65]

    Lam Tran, Thang Hoang, Alexandros Iosifidis, and Moncef Gabbouj. 2021. Multi- Model Long Short-Term Memory Network for Gait Recognition Using Window- Based Data Segment.IEEE Access9 (2021), 23833–23846. doi:10.1109/ACCESS. 2021.3057554

    arXiv 2021

  66. [66]

    Timo Von Marcard, Bodo Rosenhahn, Michael J Black, and Gerard Pons-Moll

  67. [67]

    Marvi Waheed, Hammad Afzal, and Khawir Mehmood. 2021. NT-FDS—A Noise Tolerant Fall Detection System Using Deep Learning on Wearable Devices.Sensors 21, 6 (2021), 2006. doi:10.3390/s21062006

    doi:10.3390/s21062006 2021

  68. [68]

    2013.Dynamics of Systems of Rigid Bodies

    Jens Wittenburg. 2013.Dynamics of Systems of Rigid Bodies. Vol. 33. Springer- Verlag

    2013

  69. [69]

    Huatao Xu, Pengfei Zhou, Rui Tan, and Mo Li. 2023. Practically Adopting Human Activity Recognition. InProceedings of the 29th Annual International Conference on Mobile Computing and Networking. 1–15

    2023

  70. [70]

    Huatao Xu, Pengfei Zhou, Rui Tan, Mo Li, and Guobin Shen. 2021. Limu-bert: Unleashing the potential of unlabeled data for imu sensing applications. In Proceedings of the 19th ACM Conference on Embedded Networked Sensor Systems. 220–233

    2021

  71. [71]

    Shunpei Yamaguchi, Aditya Arun, Takuya Fujiwara, Misaki Sakuta, Ryotaro Hada, Takuya Fujihashi, Takashi Watanabe, Dinesh Bharadia, and Shunsuke Saruwatari. 2024. Experience: Practical Challenges for Indoor AR Applications (ACM MobiCom ’24). Association for Computing Machinery, New York, NY, USA, 1030–1044

    2024

  72. [72]

    Sijie Yan, Yuanjun Xiong, and Dahua Lin. 2018. Spatial Temporal Graph Convolu- tional Networks for Skeleton-Based Action Recognition. arXiv:1801.07455 [cs.CV] https://arxiv.org/abs/1801.07455

    Pith/arXiv arXiv 2018

  73. [73]

    Jian Bo Yang, Minh Nhut Nguyen, Phyo Phyo San, Xiao Li Li, and Shonali Krishnaswamy. 2015. Deep convolutional neural networks on multichannel time series for human activity recognition. InProceedings of the 24th International Conference on Artificial Intelligence(Buenos Aires, Argentina)(IJCAI’15). AAAI Press, 3995–4001

    2015

  74. [74]

    Shuochao Yao, Shaohan Hu, Yiran Zhao, Aston Zhang, and Tarek Abdelzaher

  75. [75]

    Shuochao Yao, Yiran Zhao, Huajie Shao, ShengZhong Liu, Dongxin Liu, Lu Su, and Tarek Abdelzaher. 2018. Fastdeepiot: Towards understanding and optimizing neural network execution time on mobile and embedded devices. InProceedings of the 16th ACM Conference on Embedded Networked Sensor Systems. 278–291

    2018

  76. [76]

    Shuochao Yao, Yiran Zhao, Huajie Shao, Chao Zhang, Aston Zhang, Shaohan Hu, Dongxin Liu, Shengzhong Liu, Lu Su, and Tarek Abdelzaher. 2018. Sensegan: Enabling deep learning for internet of things with a semi-supervised framework. Proceedings of the ACM on interactive, mobile, wearable and ubiquitous technologies 2, 3 (2018), 1–21

    2018

  77. [77]

    Shuochao Yao, Yiran Zhao, Aston Zhang, Lu Su, and Tarek Abdelzaher. 2017. Deepiot: Compressing deep neural network structures for sensing systems with a compressor-critic framework. InProceedings of the 15th ACM conference on embedded network sensor systems. 1–14

    2017

  78. [78]

    Xinyu Yi, Yuxiao Zhou, Marc Habermann, Soshi Shimada, Vladislav Golyanik, Christian Theobalt, and Feng Xu. 2022. Physical inertial poser (pip): Physics- aware real-time human motion tracking from sparse inertial sensors. InPro- ceedings of the IEEE/CVF conference on computer vision and pattern recognition. 13167–13178

    2022

  79. [79]

    Xinyu Yi, Yuxiao Zhou, and Feng Xu. 2021. TransPose: real-time 3D human translation and pose estimation with six inertial sensors.ACM Trans. Graph.40, 4, Article 86 (July 2021), 13 pages. doi:10.1145/3450626.3459786

    arXiv 2021

  80. [80]

    Hyungjun Yoon, Hyeongheon Cha, Canh Hoang Nguyen, Taesik Gong, and Sung- Ju Lee. 2022. IMG2IMU: Applying Knowledge from Large-Scale Images to IMU Applications via Contrastive Learning.arXiv preprint arXiv:2209.00945(2022)

    Pith/arXiv arXiv 2022

Showing first 80 references.