Pith. sign in

REVIEW 5 major objections 3 minor 33 references

FLASH: Efficient Impact Fall Detection with Unified Hypergraph State-Space Model

T0 review · 5 major / 3 minor · reviewed 2026-08-01 · deepseek-v4-flash

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

Pith's one-line read FLASH detects the exact frame of a fall impact from 3D skeletons by combining a six-hyperedge biomechanical hypergraph with Mamba's linear-time state-space model, reaching 95.13% accuracy on UP-Fall and 95.83% zero-shot on UMAFall at 11.2 m

desk verdict Efficiency gain is real, but the SOTA claim is contradicted by the paper's own Table 1. read the letter →

arxiv 2607.25791 v1 pith:6JX2VARP submitted 2026-07-28 cs.CV

Tresor Y. Koffi , Youssef Mourchid , Yohan Dupuis This is my paper
classification cs.CV
keywords impactdetectionfallskeleton-basedactionrecognitionhypergraphneuralnetworksMambastate-spacemodelsmulti-jointcoordinationreal-timeinferencezero-shotgeneralization
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

This paper tries to show that the exact moment a falling person hits the ground can be detected from 3D skeleton data alone, accurately enough for real-time alerts. Existing graph networks only connect joints in pairs, but a fall impact is a whole-body event, so the authors build a hypergraph whose six hyperedges group joints into biomechanical units (torso, arms, legs, head) and combine it with Mamba's state-space model, which processes time in linear rather than quadratic complexity. On the UP-Fall dataset the model reaches 95.13% accuracy and 95.52% F1 at 11.2 ms per inference, and it transfers without fine-tuning to UMAFall at 95.83% accuracy. If these results hold, a lightweight camera- or skeleton-based system could flag falls at the moment of impact with a small, interpretable model.

What carries the argument

The load-bearing object is the normalized single incidence matrix H_norm = D_v^{-1/2} H D_e^{-1} H^T D_v^{-1/2}, built from a 33-joint × 6-hyperedge incidence matrix H. Two hypergraph convolution layers propagate features over this matrix, so a joint's representation is updated by all other joints in its biomechanical group rather than by pairwise bones alone. The spatial features are then flattened across joints and processed by a Mamba selective state-space block that runs in O(T) time, followed by three parallel temporal convolutions with kernel lengths 9, 15, and 20 to catch fine, medium, and coarse impact dynamics before frame-wise classification.

What would settle it

Run the zero-shot protocol after explicitly aligning the joint sets, coordinate frames, and impact labels of UP-Fall and UMAFall; if the 95.83% figure drops substantially once misalignment is corrected, the generalization claim is an artifact. A second check: retrain with hyperedges randomly permuted; if accuracy stays near 95%, the biomechanical grouping is not the active ingredient.

Watch

Extended reading notes

Core claim

The paper's central claim is that a single-matrix hypergraph representation, with six biomechanically motivated hyperedges grouping the 33 skeleton joints into torso, left/right arms, left/right legs, and head, is sufficient to capture the multi-joint coordination that characterizes a fall impact, and that pairing this with a Mamba selective state-space model gives frame-level impact detection that is both accurate and fast. On UP-Fall the model reports 95.13% accuracy and 95.52% F1; without any fine-tuning it transfers to UMAFall at 95.83% accuracy. The authors further claim that this single-matrix design cuts FLOPs by 71.3%, memory by 47.1%, and inference time by 93.9% relative to a dual-m

Load-bearing premise

The zero-shot generalization result assumes UP-Fall and UMAFall skeleton streams are directly comparable — matching joint sets, coordinate conventions, and impact labels — and the paper does not document how those differences are reconciled.

Editorial extensions

If this is right

  • Frame-level impact detection at 95% accuracy makes real-time alarm triggering possible from skeleton data alone, without wearable sensors.
  • The 71% FLOP reduction and 11 ms inference relative to dual-hypergraph baselines suggest the model fits on edge hardware such as phones or embedded GPUs.
  • The leg hyperedge ablation (largest drop, 2.98 pp) indicates lower-limb joint coordination is the primary signal for impact detection.
  • Occlusion robustness (only 1.38 pp drop at 30% missing joints) implies the multi-joint hyperedges provide redundancy when cameras lose track of body parts.
  • The learned attention concentrated on torso, hips, and knees gives a checkable, biomechanically sensible explanation for each prediction.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • A reader might push further: if a fixed six-hyperedge design already matches or approaches dual-matrix methods, then the extra cost of learned or dual incidence matrices may be unnecessary for other fast kinematic tasks; the same ablation logic could be applied to action spotting or surgical motion analysis.
  • The zero-shot result is the claim most worth auditing: the paper does not document how UP-Fall and UMAFall joint sets, coordinate frames, or impact labels align, so the 95.83% figure should be re-run after a documented alignment protocol.
  • One testable extension is to replace the hand-built hyperedges with a learnable grouping regularized toward biomechanics; if accuracy rises without losing interpretability, the fixed design is a floor rather than a ceiling.
  • The reported 11.2 ms is model inference only; a deployable system would face pose-estimation and preprocessing latency, so a full end-to-end latency measurement would be the natural next step.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

5 major / 3 minor

Summary. The paper proposes FLASH, a skeleton-based frame-level impact fall detection model that combines a single-matrix hypergraph representation with Mamba's selective state-space model and multi-scale temporal convolutions. The hypergraph uses six hand-designed biomechanical hyperedges over 33 joints; two hypergraph convolution layers encode spatial coordination, a Mamba block models temporal dynamics, and a multi-scale TCN with fully connected layers produces frame-wise impact classifications. Experiments are reported on UP-Fall and UMAFall, with claims of 95.13% accuracy / 95.52% F1 on UP-Fall, 95.83% zero-shot transfer to UMAFall, 37.1M parameters, 11.2 ms inference, and 71.3% FLOP reduction versus a dual-representation hypergraph baseline. The paper also claims state-of-the-art accuracy, real-time suitability, strong zero-shot generalization, and interpretable biomechanical attention.

Significance. If the claims were fully supported, the contribution would be practically valuable: a 37.1M-parameter skeleton model with 11.2 ms inference and 95.8% zero-shot transfer would be of clear interest for edge and healthcare fall detection. The efficiency-oriented design, code release, and framing around frame-level impact detection are appropriate and the efficiency numbers are plausible. However, the current validation does not support the headline 'state-of-the-art accuracy' claim: the only frame-level accuracy comparison in Table 1 shows FLASH below DistillH-Mamba, several baselines are not frame-level detectors, the zero-shot protocol is underdocumented, and the ablation does not isolate the contributions of the hypergraph and Mamba components. The paper's real potential lies in being an efficiency/accuracy trade-off study, not as submitted.

major comments (5)
  1. [Table 1 / Abstract / Conclusion] The central 'state-of-the-art accuracy' claim is directly contradicted by the paper's own numbers. In Table 1, FLASH achieves 95.13% accuracy and 95.52% F1 on UP-Fall, while DistillH-Mamba achieves 97.38% and 97.51% — a 2.25pp accuracy gap. Section 4.2 quietly downgrades the terminology to 'competitive performance,' but the Abstract and Conclusion still claim state-of-the-art performance. With no error bars or multiple-seed statistics, even 'competitive' is not statistically supported. This mismatch must be corrected; the defensible claim at present is an efficiency/accuracy trade-off, not SOTA.
  2. [Table 1 / §4.2] The comparison set is not apples-to-apples. The 'Impact' column in Table 1 shows that ST-GCN, 2s-AGCN, and Transformer are sequence-level classifiers (Impact: No); only DistillH-Mamba and FLASH perform frame-level impact localization. Sequence-level accuracy and frame-level accuracy are different metrics, so the rows for those baselines do not support the stated face-off. Baselines need to be adapted to the frame-level task, or the table should be restricted to comparable methods with clear caveats.
  3. [§4.3 / Table 2] The zero-shot cross-dataset result is load-bearing for the generalization claim, but the manuscript does not document the preprocessing alignment. To transfer from UP-Fall to UMAFall, the joint set correspondence, coordinate frame orientation, sampling rate, label definition (impact vs non-impact), and test split must be specified. If the two datasets use different skeleton conventions or label granularity, the 95.83% zero-shot accuracy may reflect preprocessing rather than learned generalization. This must be documented before the claim can be evaluated.
  4. [§4.6 / Table 4] The ablation study is incomplete. Table 4 compares full FLASH only against a dual-hypergraph baseline and a Transformer, but no arm removes the Mamba block from FLASH, and no arm removes the hypergraph block. The accuracy contributions of the hypergraph and of Mamba are therefore not isolated. Table 4 also reports only efficiency metrics, not accuracy for ablated variants. Please add hypergraph-only, Mamba-only, and full-model runs with accuracy, FLOPs, and inference time for each.
  5. [§4.4 / Fig. 2 / §3.2] The interpretability claim is partly self-fulfilling. The hypergraph H is constructed by hand as six biomechanical groups (Eq. (1), §3.2). The attention scores in Figure 2 are computed from the second hypergraph convolution layer, whose receptive field is restricted to those fixed groups. High attention on torso, knees, and hips is therefore at least partially enforced by construction, not discovered. To support the claim that the model 'automatically focuses on biomechanically relevant joints,' the authors should compare against randomly generated hyperedges or a complete-graph baseline and show that the fixed biomechanical groups outperform them in accuracy or attention quality.
minor comments (3)
  1. [§3.1, §3.2] There are numerous spacing and formatting typos, e.g., 'constructE= 6anatomical' and 'S={s 1, s2'. Please clean up the LaTeX.
  2. [§3.5] Equations (5)-(7) use kernel sizes (9,1), (15,1), (20,1), while the text says kernel sizes {9,15,20}. Please be consistent and clarify the intended tuple semantics.
  3. [References] Reference [17] is a crowd-counting paper and appears irrelevant to the claim about transformer complexity for fall detection. Consider citing a more direct source for that statement.

Circularity Check

1 steps flagged · score 2.0 of 10

Mild circularity in the interpretability claim; the main accuracy/efficiency results are empirical and not circular, though the SOTA claim is internally contradicted.

  1. self definitional [Section 3.2 (hyperedge construction) and Section 4.4 (Hypergraph Structure Analysis)]
    "Following biomechanical principles [19,20], we construct E=6 anatomical hyperedges grouping torso and lower extremity joints to capture force transmission during impact. ... The attention distribution aligns with biomechanical principles. High attention concentrates on the torso, knees, and hips primary force transmission hubs."

    The incidence matrix Hnorm (Eq. 1) is hand-built from the same biomechanical prior (torso/lower-limb groups) used later to interpret attention. Because Eq. 2 aggregates features only within those fixed hyperedges, the 'learned' concentration on torso/knees/hips is strongly constrained by construction; the interpretability result is partly a restatement of the design input rather than an independent discovery.

full rationale

FLASH's main quantitative claims (95.13% UP-Fall accuracy, 95.83% zero-shot UMAFall, 11.2 ms, 71.3% FLOP reduction) are empirical measurements on held-out test splits, not quantities fitted to define the model, so they do not reduce to their inputs by construction. The efficiency comparison uses the authors' own DistillH-Mamba [18] as a baseline and the authors' own improved UP-Fall skeletons dataset [21], but self-citation alone is not circular: the FLOP/memory/time numbers are externally measurable, and Table 1 shows FLASH is less accurate than that baseline. The one genuinely circular element is the interpretability/attention analysis, which validates biomechanical hyperedges by observing attention patterns that those hyperedges largely predetermine. Separately, but not a circularity issue, the abstract's 'state-of-the-art accuracy' is contradicted by Table 1 (DistillH-Mamba 97.38% vs FLASH 95.13%), and the zero-shot transfer assumes UP-Fall/UMAFall skeleton compatibility without documented alignment; these are consistency/validity concerns, not derivation-by-construction.

Assumptions & free parameters 2 free parameters · 4 assumptions · 0 invented entities

The central claim rests primarily on a hand-designed hyperedge topology and on two skeleton datasets (one authored by the same team), plus the unvalidated cross-dataset correspondence. No mathematical derivation or new mechanism is added beyond the existing Mamba/HGCN toolchain.

free parameters (2)
  • Hyperedge incidence matrix H (6 biomechanical groups) = Hand-designed binary matrix (E=6; exact memberships not listed)
    Defines the spatial inductive bias in Eq. (1)-(2); no sensitivity analysis; the 'discovered' joint importance in §4.4 is shaped by this choice.
  • MTCN kernel sizes = {9, 15, 20}
    Chosen to capture fine/medium/coarse temporal scales; no kernel-size ablation, so the values may be tuned to the test split.
assumptions (4)
  • standard math Mamba selective SSM equations provide valid temporal modeling with linear complexity
    Paper relies on [9] without re-derivation.
  • domain assumption The 6 hand-defined hyperedges correspond to biomechanically functional joint groups during fall impacts
    §3.2; central to the claim that hypergraph topology improves detection; the only supporting evidence is the circular attention/ablation analysis.
  • domain assumption UP-Fall 3D skeleton dataset annotations are correct ground-truth impact labels
    §4.1.1; dataset introduced by the authors in [21] with no independent validation or external benchmark split.
  • domain assumption UMAFall and UP-Fall skeletons are directly comparable for zero-shot transfer without explicit alignment of joint definitions or coordinate systems
    §4.3; no preprocessing/alignment details provided, yet zero-shot numbers are central.

how reviews work

0 comments
Cite this review

Pith. "Pith review of FLASH: Efficient Impact Fall Detection with Unified Hypergraph State-Space Model." pith.science (2026). https://pith.science/paper/6JX2VARP

@misc{pith2026260725791,
  author       = {Pith},
  title        = {Pith review of: FLASH: Efficient Impact Fall Detection with Unified Hypergraph State-Space Model},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/6JX2VARP}},
  note         = {Machine review of arXiv:2607.25791}
}
read the original abstract

Falls represent a critical public health challenge, and accurate detection of the impact moment when an individual hits the ground is crucial for timely intervention. Existing skeleton-based methods rely on graph neural networks modeling only pairwise joint connections, failing to capture multi-joint coordination characteristic of fall impacts, while transformer-based temporal models suffer from quadratic complexity limiting real-time deployment. We propose FLASH, a novel framework integrating single-matrix hypergraph representations with Mamba's selective state-space models through adaptive feedback mechanisms for efficient impact detection. Our approach constructs biomechanically-grounded hyperedges to model functional joint coordination while leveraging Mamba's linear-time complexity to capture temporal dynamics. Experiments on UP-Fall and UMAFall datasets demonstrate that FLASH achieves state-of-the-art accuracy with real-time inference capability and strong zero-shot cross-dataset generalization, while significantly reducing computational cost compared to dual-representation and transformer-based methods. The model provides interpretable feedback through learned attention patterns aligned with biomechanical principles. Code is available at https://github.com/Tresor-Koffi/FLASH-Impact-Fall-Detection.

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

33 extracted references · 3 linked inside Pith

  1. [1]

    Effective intervention requires not only detecting that a fall has occurred but also identifying the precise moment of ground impact the instant when the body hits the ground

    INTRODUCTION Falls rank as the second leading cause of unintentional injury- related deaths worldwide [1], making robust fall detection systems increasingly critical [2] [3]. Effective intervention requires not only detecting that a fall has occurred but also identifying the precise moment of ground impact the instant when the body hits the ground. This t...

  2. [2]

    We used a single-matrix hypergraph representation for human skeleton to detect impact, which encoded multi-joint coordination while avoiding dual-incidence redundancy

  3. [3]

    We integrate hypergraph spatial modeling with Mamba’s selective state-space architecture for efficient temporal processing , achieving 71.3% reduction in FLOPs and 93.9% faster inference compared to dual-representation methods, as validated through ablation studies

  4. [4]

    We provide interpretable feedback through learned joint attention patterns, demonstrating that our model automatically focuses on biomechanically relevant joints for impact detection

  5. [5]

    We validated our approach on UP-Fall [2] and UMAFall

  6. [6]

    Vision-based fall detection using ST-GCN,

    O. Keskes and R. Noumeir, “Vision-based fall detection using ST-GCN,”IEEE Access, vol. 9, pp. 28224–28236, 2021

    2021

  7. [7]

    RELATED WORK 2.1. Graph-Based Skeleton Action Recognition Graph neural networks have become the dominant approach for skeleton-based action recognition, representing joints as nodes and bones as edges [5]. Spatial-Temporal Graph Con- volutional Networks (ST-GCNs) jointly model spatial topol- ogy and temporal dynamics, achieving strong performance in both ...

  8. [8]

    Problem Formulation Given a skeletal motion sequenceS={s 1, s2,

    METHODOLOGY 3.1. Problem Formulation Given a skeletal motion sequenceS={s 1, s2, . . . , sT } where eachs t represents 3D joint positions at timet, we aim to identify the specific frames k where impact occurs. Each frame at time steptcontainsJjoints in 3D space: st ={j t 1, jt 2, . . . , jt J }wherej t i ∈R 3 represents the(x, y, z) coordinates of jointia...

Show all 33 references
  1. [9]

    Experimental Setup 4.1.1

    EXPERIMENTS AND RESULTS 4.1. Experimental Setup 4.1.1. Datasets We evaluate on two skeleton-based fall detection datasets. The 3D Skeletons UP-Fall dataset [21], built upon the orig- inal UP-Fall dataset [2], provides 3D skeleton data specifi- cally tailored for impact detecti...

  2. [10]

    arXiv:2607.25791v1 [cs.CV] 28 Jul 2026 The remainder of this paper is organized as follows: Sec- tion 2 reviews related work

    datasets, achieving state-of-the-art performance with strong cross-dataset generalization and real-time inference. arXiv:2607.25791v1 [cs.CV] 28 Jul 2026 The remainder of this paper is organized as follows: Sec- tion 2 reviews related work. Section 3 details our methodol- ogy....

    2026 arXiv

  3. [11]

    Biomechanically-grounded hyperedges capture multi-joint coordination while Mamba’s linear-time com- plexity ensures real-time efficiency

    CONCLUSION We presented FLASH, integrating single-matrix hypergraphs with Mamba’s selective state-space models for impact fall detection. Biomechanically-grounded hyperedges capture multi-joint coordination while Mamba’s linear-time com- plexity ensures real-time efficiency. F...

  4. [12]

    Fall-induced injuries and deaths among older adults,

    P. Kannus, “Fall-induced injuries and deaths among older adults,”JAMA, vol. 281, no. 20, pp. 1895–1899, 1999

    1999

  5. [13]

    UP-Fall detection dataset: A multimodal approach,

    L. Mart ´ınez-Villase˜nor, H. Ponce, J. Brieva, E. Moya- Albor, J. N ´u˜nez-Mart´ınez, and C. Pe ˜nafort-Asturiano, “UP-Fall detection dataset: A multimodal approach,” Sensors, vol. 19, no. 9, pp. 1988, 2019

    1988

  6. [14]

    Machine learning and feature ranking for impact fall detection event using multisen- sor data,

    Tresor Y Koffi, Youssef Mourchid, Mohammed Hin- dawi, and Yohan Dupuis, “Machine learning and feature ranking for impact fall detection event using multisen- sor data,” in2023 IEEE 25th International Workshop on Multimedia Signal Processing (MMSP). IEEE, 2023, pp. 1–6

    2023

  7. [15]

    Fall detec- tion based on key points of human-skeleton using Open- Pose,

    W. Chen, Z. Jiang, H. Guo, and X. Ni, “Fall detec- tion based on key points of human-skeleton using Open- Pose,”Symmetry, vol. 12, no. 5, pp. 744, 2020

    2020

  8. [16]

    Spatial temporal graph convolutional networks for skeleton-based action recog- nition,

    S. Yan, Y . Xiong, and D. Lin, “Spatial temporal graph convolutional networks for skeleton-based action recog- nition,” inProceedings of the AAAI Conference on Arti- ficial Intelligence, 2018, vol. 32, pp. 7444–7452

    2018

  9. [17]

    Hy- pergraph neural networks,

    Y . Feng, H. You, Z. Zhang, R. Ji, and Y . Gao, “Hy- pergraph neural networks,” inProceedings of the AAAI Conference on Artificial Intelligence, 2019, vol. 33, pp. 3558–3565

    2019

  10. [18]

    Transformer for skeleton-based ac- tion recognition: A review of recent advances,

    Wentian Xin, Ruyi Liu, Yi Liu, Yu Chen, Wenxin Yu, and Qiguang Miao, “Transformer for skeleton-based ac- tion recognition: A review of recent advances,”Neuro- computing, vol. 537, pp. 164–186, 2023

    2023

  11. [19]

    Mamba: Linear-time sequence modeling with selective state spaces,

    Albert Gu and Tri Dao, “Mamba: Linear-time sequence modeling with selective state spaces,” 2023

    2023

  12. [20]

    Umafall: A multisensor dataset for the research on automatic fall detection,

    Eduardo Casilari, Jose A Santoyo-Ram ´on, and Jose M Cano-Garc´ıa, “Umafall: A multisensor dataset for the research on automatic fall detection,”Procedia Com- puter Science, vol. 110, pp. 32–39, 2017

    2017

  13. [21]

    Skeleton-based fall detection with multiple inertial sensors using spatial- temporal graph convolutional networks,

    J. Yan, X. Wang, J. Shi, and S. Hu, “Skeleton-based fall detection with multiple inertial sensors using spatial- temporal graph convolutional networks,”Sensors, vol. 23, no. 4, pp. 2153, 2023

    2023

  14. [22]

    Impact detection in fall events: Leveraging spatio-temporal graph convolutional networks and recurrent neural networks using 3d skele- ton data,

    Tresor Y Koffi, Youssef Mourchid, Mohammed Hin- dawi, and Yohan Dupuis, “Impact detection in fall events: Leveraging spatio-temporal graph convolutional networks and recurrent neural networks using 3d skele- ton data,”Journal of Healthcare Informatics Research, pp. 1–37, 2025

    2025

  15. [23]

    Hypergraph neural network for skeleton-based ac- tion recognition,

    Xiaoke Hao, Jie Li, Yingchun Guo, Tao Jiang, and Ming Yu, “Hypergraph neural network for skeleton-based ac- tion recognition,”IEEE Transactions on Image Process- ing, vol. 30, pp. 2263–2275, 2021

    2021

  16. [24]

    Autoregressive adaptive hypergraph transformer for skeleton-based activity recognition,

    Abhisek Ray, Ayush Raj, and Maheshkumar H Kolekar, “Autoregressive adaptive hypergraph transformer for skeleton-based activity recognition,” in2025 IEEE/CVF Winter Conference on Applications of Computer Vision (WACV). IEEE, 2025, pp. 9690–9699

    2025

  17. [25]

    Hypergraph-based multi-view action recognition using event cameras,

    Yue Gao, Jiaxuan Lu, Siqi Li, Yipeng Li, and Shaoyi Du, “Hypergraph-based multi-view action recognition using event cameras,”IEEE Transactions on Pattern Analysis and Machine Intelligence, 2024, arXiv:2403.19316

    2024 arXiv

  18. [26]

    Ensembles of deep lstm learners for activity recognition using wearables,

    Yu Guan and Thomas Pl ¨otz, “Ensembles of deep lstm learners for activity recognition using wearables,”Pro- ceedings of the ACM on interactive, mobile, wearable and ubiquitous technologies, vol. 1, no. 2, pp. 1–28, 2017

    2017

  19. [27]

    Tcformer: A 5m- parameter transformer with density-guided aggregation for weakly-supervised crowd counting,

    Qiang Guo, Rubo Zhang, Bingbing Zhang, Jun- jie Liu, and Jianqing Liu, “Tcformer: A 5m- parameter transformer with density-guided aggregation for weakly-supervised crowd counting,”arXiv preprint arXiv:2512.22203, 2025

    2025

  20. [28]

    Distillh-mamba: A hypergraph-mamba-based knowledge distillation model for efficient impact fall detection,

    Tresor Y Koffi, Youssef Mourchid, Mohammed Hin- dawi, and Yohan Dupuis, “Distillh-mamba: A hypergraph-mamba-based knowledge distillation model for efficient impact fall detection,”IEEE Sensors Jour- nal, 2025

    2025

  21. [29]

    Finite element analysis of hip frac- ture risk in elderly female: The effects of soft tissue shape, fall direction, and interventions,

    Sotaro Murakami, Yuqing Zhao, Koji Mizuno, Minoru Yamada, Yoichi Yokoyama, Yoshitake Yamada, and Masahiro Jinzaki, “Finite element analysis of hip frac- ture risk in elderly female: The effects of soft tissue shape, fall direction, and interventions,”Journal of biomechanics, v...

    2024

  22. [30]

    The effect of fall biomechan- ics on risk for hip fracture in older adults: a cohort study of video-captured falls in long-term care,

    Yijian Yang, Vicki Komisar, Nataliya Shishov, Bryan Lo, Alexandra MB Korall, Fabio Feldman, and Stephen N Robinovitch, “The effect of fall biomechan- ics on risk for hip fracture in older adults: a cohort study of video-captured falls in long-term care,”Journal of bone and min...

    1914

  23. [31]

    An improved 3d skeletons up-fall dataset: enhancing data quality for efficient im- pact fall detection,

    Tresor Y Koffi, Youssef Mourchid, Mohammed Hin- dawi, and Yohan Dupuis, “An improved 3d skeletons up-fall dataset: enhancing data quality for efficient im- pact fall detection,” inSeventeenth International Con- ference on Machine Vision (ICMV 2024). SPIE, 2025, vol. 13517, pp. 215–222

    2024

  24. [32]

    Two-stream adaptive graph convolutional networks for skeleton- based action recognition,

    L. Shi, Y . Zhang, J. Cheng, and H. Lu, “Two-stream adaptive graph convolutional networks for skeleton- based action recognition,” inProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2019, pp. 12026–12035

    2019

  25. [33]

    Hy- pergraph transformer for skeleton-based action recogni- tion,

    Yuxuan Zhou, Zhi-Qi Cheng, Chao Li, Yanwen Fang, Yifeng Geng, Xuansong Xie, and Margret Keuper, “Hy- pergraph transformer for skeleton-based action recogni- tion,”arXiv preprint arXiv:2211.09590, 2022

    2022 arXiv

Pith tools

Reviewed August 1, 2026 · model on record in the stance chip above.