Pith. sign in

REVIEW 4 major objections 3 minor 76 references

Privacy-Preserving Multi-Stage Fall Detection Framework with Semi-supervised Federated Learning and Robotic Vision Confirmation

T0 review · 4 major / 3 minor · reviewed 2026-08-06 · deepseek-v4-flash

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

Pith's one-line read A three-stage wearable, navigation, and vision pipeline claims 99.99% fall detection accuracy by multiplying the failure rates of its three stages.

desk verdict Real system integration and a working demo, but the 99.99% headline multiplies failure rates as if the stages were parallel; the pipeline is serial, so the correct combined reliability is about 90.7%. read the letter →

arxiv 2507.10474 v1 pith:BUFJ24Y4 submitted 2025-07-14 cs.CV cs.AIcs.RO

Seyed Alireza Rahimi Azghadi , Truong-Thanh-Hung Nguyen , Helene Fournier , Monica Wachowicz , Rene Richard , Francis Palma , Hung Cao This is my paper · ORCID
classification cs.CVcs.AIcs.RO
keywords FallDetectionSemi-SupervisedFederatedLearningRoboticVisionConfirmationPrivacy-PreservingSystemsIndoorLocalizationRSSIFingerprintingEdgeComputingElderlyCare
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

The paper sets out to show that falls by older adults living alone can be detected with near-perfect reliability without keeping a camera running on them at all times. The proposed pipeline has three stages: a wearable inertial sensor and an edge device flag a possible fall from motion data (99.19% accuracy), a received-signal-strength (RSSI) indoor localization system estimates the person's position and guides a robot there (95% navigation success), and the robot's camera feed is analyzed on the edge to confirm whether someone has actually fallen (96.3% accuracy). The central claim is that the stages are complementary, so a fall escapes detection only if all three fail at once, making the combined failure rate the product $0.0081 \times 0.05 \times 0.0367 = 0.0000149$ and the overall accuracy $99.9985\%$, reported as 99.99%. This matters because the design targets two weaknesses of existing systems: false alarms from everyday movements, which the robot's visual check filters out, and privacy, since raw motion data never leaves the home and the camera activates only after a wearable alarm.

What carries the argument

The object that carries the reliability claim is the combined-failure-rate identity $P(\text{missed fall}) = \varepsilon_1 \varepsilon_2 \varepsilon_3$, evaluated in Section 6.5 as $0.0081 \times 0.05 \times 0.0367 = 0.0000149$, with combined accuracy $1 - 0.0000149 = 99.99851\%$. The supporting machinery is the semi-supervised federated fall detector SF2D, built from an LSTM autoencoder trained locally on unlabeled user data at each edge device, aggregated in the cloud with Federated Averaging, and then frozen as the encoder backbone of a classifier trained on the labeled SisFall benchmark; a BLE-RSSI fingerprinting model that estimates fall location; and a two-stage vision pipeline, YOLO object detection followed by a classifier over extracted bounding-box features, which decides whether a person is fallen. Each stage gates the next: a positive wearable detection triggers localization, which triggers robot navigation, whose camera stream is inspected by the vision model before any emergency contact is called.

What would settle it

Run the complete pipeline against a set of scripted falls on a laboratory floor and count how many produce an emergency alert. The 99.99% claim predicts about 1 undetected fall per 67,000 scripted falls (failure rate $1.49 \times 10^{-5}$), whereas the serial reading of the same stages predicts roughly 9 missed falls per 100, because a fall the wearable stage never flags never reaches the camera; even a few dozen scripted trials would separate the two predictions.

Watch

Extended reading notes

Core claim

The framework's central claim is a reliability identity: when a wearable fall detector, a robot navigation system, and a vision-based fallen-person detector are combined, the chance that a fall goes undetected is the product of the three stages' failure rates, $0.0081 \times 0.05 \times 0.0367 = 0.0000149$, which gives a combined accuracy of $99.9985\%$ and is reported as 99.99%. The wearable stage is a semi-supervised federated learning system called SF2D: an LSTM (long short-term memory) autoencoder trained on each user's unlabeled motion data at the edge is aggregated on a cloud server by Federated Averaging, and its frozen encoder feeds a classifier trained on a labeled benchmark dataset. The localization stage builds a BLE RSSI fingerprint map of the home using the robot's SLAM map, and the vision stage runs a YOLO (single-pass real-time object detector) that distinguishes fallen people from people resting or sitting, with the best configuration reaching 96.3% accuracy. The paper argues that this architecture is both accurate and private: raw user data never leaves the edge device, and visual inspection is triggered only after a potential fall is detected.

Load-bearing premise

The load-bearing premise is that the three stages fail independently of each other, so that a fall is missed only when the wearable detector, the robot navigation, and the camera check all fail together; the paper's headline number multiplies the three failure rates on that basis, even though its own workflow description has each stage start only after the previous one succeeds.

Editorial extensions

If this is right

  • If the combined-reliability claim holds, a genuine fall fails to produce an alert only when the wearable detector, the navigation, and the vision check fail simultaneously, an event the paper estimates at 0.0015% of cases.
  • False alarms, the main weakness of wearable-only detectors, are filtered out by the robot's visit: when the wearable flags an everyday movement as a fall, the vision stage sees a person who has not fallen, the robot returns to its station, and the episode is stored as a misclassification for future retraining.
  • Privacy is structural rather than added on: the cloud receives only aggregated model updates, never raw motion data, and the robot's camera turns on only after an initial fall alarm, so no continuous video record of the home exists.
  • Deployment does not require labeled data from each individual user: a federated encoder trained on unlabeled data plus a classifier trained on a labeled benchmark is enough, which the paper argues makes the approach practical for real homes.

Reading between the lines

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

  • The paper's formula measures the probability that at least one stage reacts rather than the probability that a genuine fall is confirmed: a wearable false alarm that sends the robot out and is then rejected by vision counts as a success in the product formula but triggers no emergency response, so 'accuracy' in the abstract and 'detection of real falls' are different quantities.
  • The 99.99% figure treats the three stages as redundant paths with independent failures, while the paper's own workflow is largely serial, since the camera never runs if the wearable detector misses the fall; under a serial reading the end-to-end detection rate of a real fall is the product of the success rates, $0.9919 \times 0.95 \times 0.963 \approx 90.7\%$.
  • An end-to-end field trial with scripted falls, counting how many produce an emergency alert, is the test the laboratory demonstration does not yet provide, and the authors' stated plan for long-term studies with older adults could settle whether the parallel or serial model describes the deployed system.
  • The design pattern of using a cheap always-on wearable signal to gate an expensive privacy-intrusive confirmation generalizes to other assisted-living alarms, such as seizure detection or wandering alerts, where a camera or robot is activated only when a low-dimensional sensor indicates an emergency.
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

4 major / 3 minor

Summary. The paper proposes a three-stage fall detection framework: (1) a semi-supervised federated learning classifier on wearable IMU data (SF2D), (2) an RSSI-based localization and robot navigation subsystem, and (3) a vision-based fallen-person confirmation model. The authors report component accuracies of 99.19%, 95%, and 96.3%, and combine their failure rates to claim an overall accuracy of 99.99%. The manuscript also documents a hardware prototype, offline training pipelines on public benchmarks (SisFall and E-FPDS), and an end-to-end laboratory demonstration using ROS2 and Nav2.

Significance. If the 99.99% overall accuracy were valid, this would be a notable contribution to privacy-preserving multi-stage fall detection, combining federated semi-supervised learning, mobile robotics, and vision confirmation in a single deployed prototype. The paper has concrete strengths: a real hardware implementation, re-run experiments on public datasets, detailed documentation of the ROS2/Nav2 integration, and an explicit privacy-preserving design that sends video to the edge only after a wearable alert. However, the headline reliability claim is computed with an incorrect probabilistic composition of serial stages, and one of the three component rates is taken from external documentation rather than measured in the authors' deployment. Recomputing the serial pipeline gives roughly 90.7% end-to-end detection of real falls, which substantially weakens the claimed contribution. The system concept remains interesting, but the central quantitative claim is not supported.

major comments (4)
  1. [Section 6.5, 'Quantitative Evaluation'] The combined accuracy is computed as 1 - (0.0081 × 0.05 × 0.0367) = 99.9985%, which is valid only if the three stages are independent redundant paths that fail jointly. The workflow in Section 3.1 and the text in Section 6.5 ('the success of one component dependent on the preceding one') describe a serial pipeline: a false negative at SF2D prevents localization, navigation, and vision confirmation from ever being triggered. The correct serial combination is 0.9919 × 0.95 × 0.963 ≈ 0.907, i.e., about 90.7% end-to-end detection of real falls, not 99.99%. The abstract, Section 6.5, and Section 8 repeat the 99.99% claim and must be revised.
  2. [Section 6.5, 'Nav2 + SLAM Toolbox Performance'] The 95% navigation success rate is not measured in this study; it is cited from Nav2 documentation and developer evaluations (Macenski et al., 2020; Macenski and Jambrecic, 2021). No navigation success count from the end-to-end laboratory experiment is reported, so the 0.05 failure rate used in the headline calculation is an external benchmark figure, not a validated property of this robot, map, or environment. The end-to-end experiment should report its own navigation success rate, or the quantitative claim should be restricted to measured components.
  3. [Section 6.4 and Table 12] The 96.3% vision accuracy is the best value among 16 YOLO variants crossed with four classifiers, selected after inspecting test-set performance. The paper does not describe a validation-based model-selection procedure or any correction for multiple comparisons, so this point estimate likely overstates expected accuracy on a fresh deployment. The headline calculation uses this optimistic value without confidence intervals or error propagation.
  4. [Section 6.5, 'End-to-End Evaluation'] The 'end-to-end evaluation' is qualitative: Figures 6–8 show navigation and detection screenshots, but no end-to-end success rate, latency, or false-alarm count is reported. The 99.9985% figure is therefore not an experimental result but an arithmetic product of component accuracies, one of which is externally sourced. A claim of high reliability requires either a measured end-to-end rate or a properly propagated estimate with uncertainty.
minor comments (3)
  1. [Section 6.4.1 and Table 10] The text states that YOLOv10n 'has the highest mAP50 score (0.828),' but Table 10 reports yolo11n with mAP50 = 0.842, which is higher; the sentence is inconsistent with the table.
  2. [Abstract, Section 6.5, Section 8] The numerical value is reported as 99.9985% in the Section 6.5 calculation but as 99.99% in the abstract and conclusion; rounding should be stated explicitly, and the corrected serial calculation will change both values.
  3. [Section 5.3.3 and Section 6.3] The term 'Mean Distance Error 13' appears without an equation number at first mention; Equation (13) is only introduced later in Section 6.3, which makes the reference unclear.

Circularity Check

0 steps flagged · score 2.0 of 10

No significant circularity: the 99.99% figure is an algebraic composition of independently measured component accuracies, though the serial/parallel reliability calculus is incorrect.

full rationale

The headline accuracy is not fitted to a target or defined circularly in terms of itself. The SF2D 99.19% figure is measured on the public SisFall benchmark (Sections 5.2 and 6.1, Table 6), the vision 96.3% figure is the best test-set result on the public E-FPDS dataset (Section 6.4, Table 12), and the localization model is trained on newly collected BLE fingerprint data in this work (Sections 5.3 and 6.3). The combined figure is computed as 1 - (0.0081 x 0.05 x 0.0367) = 99.99851% in Section 6.5; that is an algebraic identity once the three component rates are accepted, so the output is not masking a hidden fit to itself. The main problem is a soundness defect, not circularity: the paper multiplies failure rates as though the stages were parallel redundant detectors, while its own workflow (Section 3.1) is serial, so a false negative at the first stage never reaches the robot. That is an incorrect reliability model, but it does not reduce the derivation to its inputs in the circularity sense. Self-citations exist (Azghadi et al. 2024 and 2025 for the SF2D and RSSI-fingerprinting methods; Nguyen et al. 2025 for contestable-AI framing), but the experiments are re-run here on external benchmarks and those self-citations are not the load-bearing evidence for the 99.99% claim. Accordingly, no circular step is identified; the low score reflects only minor, non-load-bearing self-citation.

Assumptions & free parameters 4 free parameters · 3 assumptions · 0 invented entities

The central claim of 99.99% accuracy rests on three inputs: the SF2D failure rate, a navigation success rate borrowed from Nav2 documentation, and a vision-confirmation accuracy selected on the test set. The free parameters are these input rates and the preprocessing choices that produce them; no new physical entities are postulated.

free parameters (4)
  • SF2D failure rate = 0.0081
    Derived from the reported FL accuracy of 99.19% (Table 6), used as an input to the combined reliability product in Section 6.5.
  • Navigation success rate = 95%
    Used in the combined failure-rate product (Section 6.5), but not measured in this deployment; taken from Nav2 documentation (Macenski et al., 2020; Macenski and Jambrecic, 2021).
  • Vision confirmation failure rate = 0.0367
    Derived from the best test-set accuracy of 96.3% (Table 12, yolo11l + RandomForest), selected after evaluating 16 YOLO models and 4 classifiers on the same test set.
  • BLE missing RSSI imputation = -100 dBm
    Missing values in the fingerprinting dataset are replaced with the minimum signal value, a modeling choice in Sections 4.2 and 5.3.2 that affects the localization model inputs.
assumptions (3)
  • ad hoc to paper Component failures are independent and the system succeeds unless all stages fail
    Section 6.5 multiplies per-stage failure rates to get the overall failure rate, treating the pipeline as parallel redundancy even though the text states that the success of each component depends on the preceding one.
  • domain assumption Nav2 navigation success rate of 95% transfers from documentation to the experimental lab environment
    Section 6.5 cites Macenski et al. benchmark reports instead of measuring navigation success in this deployment.
  • domain assumption SisFall and FPDS datasets are representative of real-world falls and home environments
    Sections 5.2.1 and 5.4.1 use these public datasets as proxies for real older-adult falls; no field validation with older participants is reported.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Privacy-Preserving Multi-Stage Fall Detection Framework with Semi-supervised Federated Learning and Robotic Vision Confirmation." pith.science (2026). https://pith.science/paper/BUFJ24Y4

@misc{pith2026250710474,
  author       = {Pith},
  title        = {Pith review of: Privacy-Preserving Multi-Stage Fall Detection Framework with Semi-supervised Federated Learning and Robotic Vision Confirmation},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/BUFJ24Y4}},
  note         = {Machine review of arXiv:2507.10474}
}
read the original abstract

The aging population is growing rapidly, and so is the danger of falls in older adults. A major cause of injury is falling, and detection in time can greatly save medical expenses and recovery time. However, to provide timely intervention and avoid unnecessary alarms, detection systems must be effective and reliable while addressing privacy concerns regarding the user. In this work, we propose a framework for detecting falls using several complementary systems: a semi-supervised federated learning-based fall detection system (SF2D), an indoor localization and navigation system, and a vision-based human fall recognition system. A wearable device and an edge device identify a fall scenario in the first system. On top of that, the second system uses an indoor localization technique first to localize the fall location and then navigate a robot to inspect the scenario. A vision-based detection system running on an edge device with a mounted camera on a robot is used to recognize fallen people. Each of the systems of this proposed framework achieves different accuracy rates. Specifically, the SF2D has a 0.81% failure rate equivalent to 99.19% accuracy, while the vision-based fallen people detection achieves 96.3% accuracy. However, when we combine the accuracy of these two systems with the accuracy of the navigation system (95% success rate), our proposed framework creates a highly reliable performance for fall detection, with an overall accuracy of 99.99%. Not only is the proposed framework safe for older adults, but it is also a privacy-preserving solution for detecting falls.

Figures

Figures reproduced from arXiv: 2507.10474 by the authors.

Figure 1
Figure 1. Our Fall Detection Framework Workflow 3.2. Components of Framework This section describes the various entities within our frame￾work and their interactions. As illustrated in [PITH_FULL_IMAGE:figures/full_fig_p009_1.png] view at source ↗
Figure 2
Figure 2. Fall Detection Components and Systems of the Proposed Framework [PITH_FULL_IMAGE:figures/full_fig_p010_2.png] view at source ↗
Figure 3
Figure 3. Components figure 5.1.1. Cloud Server A high-performance computing system located in our office was used as the cloud server in our architecture. This machine is only connected to the university’s internal network and is not equipped with any wireless communication modules. The sys￾tem is a Dell Precision 3660 workstation, equipped with a 13th Gen Intel® CoreTM i9-13900K processor based on the x86 64 architecture, f… view at source ↗
Figures from the paper (12 more)
Figure 4
Figure 4. Figure 4: RViz application during the data collection process [PITH_FULL_IMAGE:figures/full_fig_p019_4.png]
Figure 5
Figure 5. Figure 5: Overlapping the signal strength from one anchor with the sample [PITH_FULL_IMAGE:figures/full_fig_p021_5.png]
Figure 6
Figure 6. Figure 6: Three figures from the navigation process from the origin to the loca [PITH_FULL_IMAGE:figures/full_fig_p024_6.png]
Figure 8
Figure 8. Figure 8: Two extra figures for providing more information [PITH_FULL_IMAGE:figures/full_fig_p025_8.png]
Figure 7
Figure 7. Figure 7: Three figures from the navigation process from the origin to the loca [PITH_FULL_IMAGE:figures/full_fig_p025_7.png]
Figure 10
Figure 10. Figure 10: mAP50 vs Inference time • mAP50 vs. Model Size [PITH_FULL_IMAGE:figures/full_fig_p029_10.png]
Figure 9
Figure 9. Figure 9: mAP50 vs Training time vs Size • mAP50 vs. Inference Time [PITH_FULL_IMAGE:figures/full_fig_p029_9.png]
Figure 11
Figure 11. Figure 11: mAP50 vs Model Size such as model size and inference time [PITH_FULL_IMAGE:figures/full_fig_p030_11.png]
Figure 12
Figure 12. Figure 12: Trained Yolo models sorted based on MAP50 metric [PITH_FULL_IMAGE:figures/full_fig_p030_12.png]
Figure 13
Figure 13. Figure 13: Detailed information on the performance of the best classifiers [PITH_FULL_IMAGE:figures/full_fig_p033_13.png]
Figure 14
Figure 14. Figure 14: Performance Comparison of Average YOLO Models Metrics For [PITH_FULL_IMAGE:figures/full_fig_p034_14.png]
Figure 15
Figure 15. Figure 15: Performance Comparison of Average Classifiers Models Metrics For [PITH_FULL_IMAGE:figures/full_fig_p034_15.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

76 extracted references · 67 canonical work pages

  1. [1]

    write newline

    " write newline "" before.all 'output.state := FUNCTION n.dashify 't := "" t empty not t #1 #1 substring "-" = t #1 #2 substring "--" = not "--" * t #2 global.max substring 't := t #1 #1 substring "-" = "-" * t #2 global.max substring 't := while if t #1 #1 substring * t #2 global.max substring 't := if while FUNCTION word.in bbl.in ":" * " " * FUNCTION f...

  2. [2]

    write newline

    " write newline "" before.all 'output.state := FUNCTION n.dashify 't := "" t empty not t #1 #1 substring "-" = t #1 #2 substring "--" = not "--" * t #2 global.max substring 't := t #1 #1 substring "-" = "-" * t #2 global.max substring 't := while if t #1 #1 substring * t #2 global.max substring 't := if while FUNCTION word.in bbl.in ":" * " " * FUNCTION f...

  3. [3]

    , author Natsheh, E

    author Abu Kharmeh, S. , author Natsheh, E. , author Sulaiman, B. , author Abuabiah, M. , author Tarapiah, S. , year 2023 . title Indoor WiFi -beacon dataset construction using autonomous low-cost robot for 3d location estimation . journal Applied Sciences volume 13 , pages 6768 . https://www.mdpi.com/2076-3417/13/11/6768, :10.3390/app13116768. note numbe...

    work page doi:10.3390/app13116768 2023

  4. [4]

    , year 2025

    author Ajibade, B.O. , year 2025 . title Falls risk assessment and prevention in older people in healthcare facilities . journal British Journal of Nursing volume 34 , pages 381--385

    work page 2025

  5. [5]

    , author Keller, I

    author Alfrink, K. , author Keller, I. , author Kortuem, G. , author Doorn, N. , year 2023 . title Contestable ai by design: Towards a framework . journal Minds and Machines volume 33 , pages 613--639

    work page 2023

  6. [6]

    , author Mih, A.N

    author Azghadi, S.A.R. , author Mih, A.N. , author Kawnine, A. , author Wachowicz, M. , author Palma, F. , author Cao, H. , year 2024 . title An adaptive indoor localization approach using wifi rssi fingerprinting with slam-enabled robotic platform and deep neural networks , in: booktitle 2024 34th International Conference on Collaborative Advances in Sof...

    work page 2024

  7. [7]

    , author Thanh, H.T.N

    author Azghadi, S.A.R. , author Thanh, H.T.N. , author Kondratova, I. , author Fournier, H. , author Wachowicz, M. , author Palma, F. , author Richard, R. , author Cao, H. , year 2025 . title Sf2d: Semi-supervised Federated Learning for Fall Detection using ( Un )labelled Data in Edge - Cloud . journal Proceedings of the Canadian Conference on Artificial ...

    work page 2025

  8. [8]

    , author Officer, A

    author Beard, J.R. , author Officer, A. , author De Carvalho, I.A. , author Sadana, R. , author Pot, A.M. , author Michel, J.P. , author Lloyd-Sherlock, P. , author Epping-Jordan, J.E. , author Peeters, G.G. , author Mahanani, W.R. , et al., year 2016 . title The world report on ageing and health: a policy framework for healthy ageing . journal The lancet...

    work page 2016

Show all 76 references
  1. [9]

    , et al., year 2020

    author Beutel, D.J. , et al., year 2020 . title Flower: A friendly federated learning research framework . journal arXiv preprint arXiv:2007.14390

    2020 arXiv

  2. [10]

    title Directive on automated decision-making - canada.ca

    author Canada , year 2019 . title Directive on automated decision-making - canada.ca . https://www.tbs-sct.canada.ca/pol/doc-eng.aspx?id=32592

    2019

  3. [11]

    , year 2025

    author Canada, H. , year 2025 . title Pre-market guidance for machine learning-enabled medical devices

    2025

  4. [12]

    , author Sapienza, S

    author Capra, M. , author Sapienza, S. , author Motto Ros, P. , author Serrani, A. , author Martina, M. , author Puiatti, A. , author Bonato, P. , author Demarchi, D. , year 2020 . title Assessing the feasibility of augmenting fall detection systems by relying on uwb-based pos...

    2020

  5. [13]

    , author Thompson, H

    author Chaudhuri, S. , author Thompson, H. , author Demiris, G. , year 2014 . title Fall detection devices and their use with older adults: a systematic review . journal Journal of geriatric physical therapy volume 37 , pages 178--196

    2014

  6. [14]

    , author Labrador, M.A

    author Delahoz, Y.S. , author Labrador, M.A. , year 2014 . title Survey on fall detection and fall prevention using wearable and external sensors . journal Sensors volume 14 , pages 19806--19842

    2014

  7. [15]

    , et al., year 2022

    author Diao, E. , et al., year 2022 . title Semifl: Semi-supervised federated learning for unlabeled clients with alternate training . journal Advances in Neural Information Processing Systems volume 35 , pages 17871--17884

    2022

  8. [16]

    , author Pham, M

    author Do, H.M. , author Pham, M. , author Sheng, W. , author Yang, D. , author Liu, M. , year 2018 . title Rish: A robot-integrated smart home for elderly care . journal Robotics and Autonomous Systems volume 101 , pages 74--92

    2018

  9. [17]

    , author Khazri, A

    author Elaoud, A. , author Khazri, A. , author Barhoumi, W. , year 2024 . title An explainable method for cost-efficient multi-view fall detection , in: booktitle 2024 IEEE 27th International Symposium on Real-Time Distributed Computing (ISORC) , organization IEEE . pp. pages 1--10

    2024

  10. [18]

    , author Abdellatif, A

    author Elwaly, A. , author Abdellatif, A. , author El-Shaer, Y. , year 2024 . title New eldercare robot with path-planning and fall-detection capabilities . journal Applied Sciences volume 14 , pages 2374

    2024

  11. [19]

    , year 2025

    author England, N. , year 2025 . title Nationwide roll out of artificial intelligence tool that predicts falls and viruses

    2025

  12. [20]

    , author Krishaa, L

    author Foo, C.D. , author Krishaa, L. , author Tang Si Ik, B. , author Kosycarz, E. , author Wang, J. , author Teo, K.W. , year 2025 . title Ai and digital technology paradigm for seniors: a singapore lens to healthy longevity . journal Frontiers in Public Health volume 12 , p...

    2025

  13. [21]

    , author Manresa-Yee, C

    author Gaya-Morey, F.X. , author Manresa-Yee, C. , author Buades-Rubio, J.M. , year 2024 . title Deep learning for computer vision based activity recognition and fall detection of the elderly: a systematic review . journal Applied Intelligence volume 54 , pages 8982--9007

    2024

  14. [22]

    , author Ghosh, S.K

    author Ghosh, S. , author Ghosh, S.K. , year 2023 . title Feel: Federated learning framework for elderly healthcare using edge-iomt . journal IEEE Transactions on Computational Social Systems volume 10 , pages 1800--1809

    2023

  15. [23]

    , year 2009

    author Giorgino, T. , year 2009 . title Computing and visualizing dynamic time warping alignments in r: the dtw package . journal Journal of statistical Software volume 31 , pages 1--24

    2009

  16. [24]

    , author Rodr \' guez, V

    author Guti \'e rrez, J. , author Rodr \' guez, V. , author Martin, S. , year 2021 . title Comprehensive review of vision-based fall detection systems . journal Sensors volume 21 , pages 947

    2021

  17. [25]

    , author Venkatesh, T

    author Hardani, S. , author Venkatesh, T. , year 2020 . title Autumn leaves: Automatic fall detection, alert and assessment to improve pre-hospital emergency care response

    2020

  18. [26]

    , author Schmidhuber, J

    author Hochreiter, S. , author Schmidhuber, J. , year 1997 . title Long short-term memory . journal Neural computation volume 9 , pages 1735--1780

    1997

  19. [27]

    , author Florence, C

    author Houry, D. , author Florence, C. , author Baldwin, G. , author Stevens, J. , author McClure, R. , year 2016 . title The cdc injury center’s response to the growing public health problem of falls among older adults . journal American journal of lifestyle medicine volume 1...

    2016

  20. [28]

    , year 1986

    author Hunter, J.S. , year 1986 . title The exponentially weighted moving average . journal Journal of quality technology volume 18 , pages 203--210

    1986

  21. [29]

    , author Medrano, C

    author Igual, R. , author Medrano, C. , author Plaza, I. , year 2013 . title Challenges, issues and trends in fall detection systems . journal Biomedical engineering online volume 12 , pages 66

    2013

  22. [30]

    , author Manikandan, V

    author Inturi, A.R. , author Manikandan, V. , author Hu, Y.C. , year 2025 . title Technical insights into vision-based fall detection systems: performances, challenges, and constraints . journal AI & SOCIETY , pages 1--13

    2025

  23. [31]

    , et al., year 2020

    author James, S.L. , et al., year 2020 . title The global burden of falls: global, regional and national estimates of morbidity and mortality from the global burden of disease study 2017 . journal Injury prevention volume 26 , pages i3--i11

    2020

  24. [32]

    , et al., year 2022

    author Ji, S. , et al., year 2022 . title Sifall: Practical online fall detection with rf sensing , in: booktitle Proceedings of the 20th ACM Conference on Embedded Networked Sensor Systems , pp. pages 563--577

    2022

  25. [33]

    , author Sirithunge, C

    author Kalinga, T. , author Sirithunge, C. , author Buddhika, A. , author Jayasekara, P. , author Perera, I. , year 2020 . title A fall detection and emergency notification system for elderly , in: booktitle 2020 6th international conference on control, automation and robotics...

    2020

  26. [34]

    , author Shehata, H.I

    author Karar, M.E. , author Shehata, H.I. , author Reyad, O. , year 2022 . title A survey of iot-based fall detection for aiding elderly care: Sensors, methods, challenges and future trends . journal Applied Sciences volume 12 , pages 3276

    2022

  27. [35]

    , author Hernandez, R

    author Liang, F. , author Hernandez, R. , author Lu, J. , author Ong, B. , author Moore, M.J. , author Sheng, W. , author Zhang, S. , year 2021 . title Collaborative fall detection using a wearable device and a companion robot , in: booktitle 2021 IEEE International Conference...

    2021

  28. [36]

    , author Velloso, E

    author Lyons, H. , author Velloso, E. , author Miller, T. , year 2021 . title Conceptualising contestability: Perspectives on contesting algorithmic decisions . journal Proceedings of the ACM on Human-Computer Interaction volume 5 , pages 1--25

    2021

  29. [37]

    , author Foote, T

    author Macenski, S. , author Foote, T. , author Gerkey, B. , author Lalancette, C. , author Woodall, W. , year 2022 . title Robot operating system 2: Design, architecture, and uses in the wild . journal Science Robotics volume 7 , pages eabm6074 . https://www.science.org/doi/a...

    2022 doi

  30. [38]

    , author Jambrecic, I

    author Macenski, S. , author Jambrecic, I. , year 2021 . title SLAM Toolbox : SLAM for the dynamic world . journal Journal of Open Source Software volume 6 , pages 2783 . https://joss.theoj.org/papers/10.21105/joss.02783, :10.21105/joss.02783

    2021 doi

  31. [39]

    , author Mart \' n, F

    author Macenski, S. , author Mart \' n, F. , author White, R. , author Clavero, J.G. , year 2020 . title The marathon 2: A navigation system , in: booktitle 2020 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS) , organization IEEE . pp. pages 2718--2725

    2020

  32. [40]

    , author Iglesias-Iglesias, C

    author Maldonado-Bascon, S. , author Iglesias-Iglesias, C. , author Mart \' n-Mart \' n, P. , author Lafuente-Arroyo, S. , year 2019 . title Fallen people detection capabilities using assistive robot . journal Electronics volume 8 , pages 915

    2019

  33. [41]

    , author Iglesias-Iglesias, C

    author Maldonado-Bascón, S. , author Iglesias-Iglesias, C. , author Martín-Martín, P. , author Lafuente-Arroyo, S. , year 2019-09 . title Fallen people detection capabilities using assistive robot . journal Electronics volume 8 , pages 915 . https://www.mdpi.com/2079-9292/8/9/...

    2019 doi

  34. [42]

    , author Ponce, H

    author Mart \' nez-Villase \ n or, L. , author Ponce, H. , author Brieva, J. , author Moya-Albor, E. , author N \'u \ n ez-Mart \' nez, J. , author Pe \ n afort-Asturiano, C. , year 2019 . title Up-fall detection dataset: A multimodal approach . journal Sensors volume 19 , pages 1988

    2019

  35. [43]

    , author Trybulska, A

    author Mikos, M. , author Trybulska, A. , author Czerw, A. , year 2021 . title Falls-the socio-economic and medical aspects important for developing prevention and treatment strategies . journal Annals of agricultural and environmental medicine volume 28

    2021

  36. [44]

    , year 2022

    author Neuwirth, R. , year 2022 . title The eu artificial intelligence act . journal The EU Artificial Intelligence Act volume 106

    2022

  37. [45]

    , author Rahimi, A

    author Nguyen, H. , author Rahimi, A. , author Whitford, V. , author Fournier, H. , author Kondratova, I. , author Richard, R. , author Cao, H. , year 2025 . title Heart2mind: Human-centered contestable psychiatric disorder diagnosis system using wearable ecg monitors . journa...

    2025 arXiv

  38. [46]

    , author Tang, H

    author Ni, J. , author Tang, H. , author Haque, S.T. , author Yan, Y. , author Ngu, A.H. , year 2024 . title A survey on multimodal wearable sensor-based human action recognition . journal arXiv preprint arXiv:2404.15349

    2024 arXiv

  39. [47]

    , author Barshan, B

    author \"O zdemir, A.T. , author Barshan, B. , year 2014 . title Detecting falls with wearable sensors using machine learning techniques . journal Sensors volume 14 , pages 10691--10708

    2014

  40. [48]

    , author Rosberg, P.C

    author Patricia, A.C.P. , author Rosberg, P.C. , author Butt-Aziz, S. , author Alberto, P.M.M. , author Roberto-Cesar, M.O. , author Miguel, U.T. , author Naz, S. , year 2024 . title Semi-supervised ensemble learning for human activity recognition in casas kyoto dataset . jour...

    2024

  41. [49]

    , author Chiaro, D

    author Qi, P. , author Chiaro, D. , author Piccialli, F. , year 2023 . title Fl-fd: Federated learning-based fall detection with multimodal data fusion . journal Information fusion volume 99 , pages 101890

    2023

  42. [50]

    , author Karuppiah, A

    author Ramachandran, A. , author Karuppiah, A. , year 2020 . title A survey on recent advances in wearable fall detection systems . journal BioMed research international volume 2020 , pages 2167160

    2020

  43. [51]

    , author Divvala, S

    author Redmon, J. , author Divvala, S. , author Girshick, R. , author Farhadi, A. , year 2016 . title You only look once: Unified, real-time object detection , in: booktitle Proceedings of the IEEE conference on computer vision and pattern recognition , pp. pages 779--788

    2016

  44. [52]

    , year 2016

    author Regulation, P. , year 2016 . title Regulation (eu) 2016/679 of the european parliament and of the council . journal Regulation (eu) volume 679 , pages 2016

    2016

  45. [53]

    , year 2023

    author Research, G.V. , year 2023 . title Fall detection systems market size, share, & trends analysis report by product, by system (wearable, non wearable), by technology, by component, by end use, by region, and segment forecasts, 2024 - 2030 . https://www.grandviewresearch....

    2023

  46. [54]

    , author Yamaguchi, H

    author Rizk, H. , author Yamaguchi, H. , author Youssef, M. , author Higashino, T. , year 2023 . title Laser range scanners for enabling zero-overhead WiFi -based indoor localization system . journal ACM Transactions on Spatial Algorithms and Systems volume 9 , pages 4:1--4:25...

    2023 doi

  47. [55]

    , et al., year 2023

    author Sathya, P. , et al., year 2023 . title Federated learning based elderly fall detection using edge computing , in: booktitle 2023 International Conference on Next Generation Electronics (NEleX) , organization IEEE . pp. pages 1--5

    2023

  48. [56]

    , author Golay, M.J

    author Savitzky, A. , author Golay, M.J. , year 1964 . title Smoothing and differentiation of data by simplified least squares procedures. journal Analytical chemistry volume 36 , pages 1627--1639

    1964

  49. [57]

    , author Erfurth, C

    author Schmidt, L. , author Erfurth, C. , year 2024 . title The future of ageing: The impact of smart home technologies on ageing in place , in: booktitle International Conference on Innovations for Community Services , organization Springer . pp. pages 83--101

    2024

  50. [58]

    , author Pendão, C

    author Silva, I. , author Pendão, C. , author Torres-Sospedra, J. , author Moreira, A. , year 2023 . title Industrial environment multi-sensor dataset for vehicle indoor tracking with wi-fi, inertial and odometry data . journal Data volume 8 , pages 157 . https://www.mdpi.com/...

    2023 doi

  51. [59]

    , author Chatterjee, K

    author Singh, A. , author Chatterjee, K. , year 2021 . title Securing smart healthcare system with edge computing . journal Computers & Security volume 108 , pages 102353

    2021

  52. [60]

    , author Kumar, S.S

    author Sona, K. , author Kumar, S.S. , year 2024 . title Iot-based system for real-time fall risk assessment and health monitoring . journal Journal of Electronics and Electrical Engineering , pages 514--529

    2024

  53. [61]

    , author Vlassov, V

    author Sozinov, K. , author Vlassov, V. , author Girdzijauskas, S. , year 2018 . title Human activity recognition using federated learning , in: booktitle 2018 IEEE Intl Conf on Parallel & Distributed Processing with Applications, Ubiquitous Computing & Communications, Big Dat...

    2018

  54. [62]

    , author L \'o pez, J.D

    author Sucerquia, A. , author L \'o pez, J.D. , author Vargas-Bonilla, J.F. , year 2017 . title Sisfall: A fall and movement dataset . journal Sensors volume 17 , pages 198

    2017

  55. [63]

    , author Matsubara, Y

    author Sumiya, T. , author Matsubara, Y. , author Nakano, M. , author Sugaya, M. , year 2015 . title A mobile robot for fall detection for elderly-care . journal Procedia computer science volume 60 , pages 870--880

    2015

  56. [64]

    , author Taylor, W

    author Tahir, A. , author Taylor, W. , author Taha, A. , author Usman, M. , author Shah, S.A. , author Imran, M.A. , author Abbasi, Q.H. , year 2022 . title Iot based fall detection system for elderly healthcare , in: booktitle Internet of Things for Human-Centered Design: App...

    2022

  57. [65]

    , et al., year 2023

    author Tashakori, A. , et al., year 2023 . title Semipfl: Personalized semi-supervised federated learning framework for edge intelligence . journal IEEE Internet of Things Journal volume 10 , pages 9161--9176

    2023

  58. [66]

    , author Hoos, H.H

    author Van Engelen, J.E. , author Hoos, H.H. , year 2020 . title A survey on semi-supervised learning . journal Machine learning volume 109 , pages 373--440

    2020

  59. [67]

    , author Ellul, J

    author Wang, X. , author Ellul, J. , author Azzopardi, G. , year 2020 . title Elderly fall detection systems: A literature survey . journal Frontiers in Robotics and AI volume 7 , pages 71

    2020

  60. [68]

    , author Chen, X

    author Wu, Q. , author Chen, X. , author Zhou, Z. , author Zhang, J. , year 2020 a. title Fedhome: Cloud-edge based personalized federated learning for in-home health monitoring . journal IEEE Transactions on Mobile Computing volume 21 , pages 2818--2832

    2020

  61. [69]

    , author Gu, Y

    author Wu, T. , author Gu, Y. , author Chen, Y. , author Wang, J. , author Zhang, S. , year 2020 b. title A mobile cloud collaboration fall detection system based on ensemble learning , in: booktitle Proceedings of the 22nd International ACM SIGACCESS Conference on Computers a...

    2020

  62. [70]

    , author Tsanousa, A

    author Xefteris, V.R. , author Tsanousa, A. , author Meditskos, G. , author Vrochidis, S. , author Kompatsiaris, I. , year 2021 . title Performance, challenges, and limitations in multimodal fall detection systems: A review . journal IEEE Sensors Journal volume 21 , pages 18398--18409

    2021

  63. [71]

    , et al., year 2023

    author Xiao, H. , et al., year 2023 . title Towards privacy-supporting fall detection via deep unsupervised rgb2depth adaptation . journal IEEE Sensors Journal

    2023

  64. [72]

    title ROSMASTER x3 ROS 2 robot with mecanum wheel for jetson NANO 4gb/orin NANO /orin NX / RaspberryPi 4b

    author Yahboom , year 2024 . title ROSMASTER x3 ROS 2 robot with mecanum wheel for jetson NANO 4gb/orin NANO /orin NX / RaspberryPi 4b . https://category.yahboom.net/products/rosmaster-x3

    2024

  65. [73]

    , author Chen, Z

    author Yu, H. , author Chen, Z. , author Zhang, X. , author Chen, X. , author Zhuang, F. , author Xiong, H. , author Cheng, X. , year 2021 . title Fedhar: Semi-supervised online learning for personalized federated human activity recognition . journal IEEE transactions on mobil...

    2021

  66. [74]

    , et al., year 2022

    author Yu, Z. , et al., year 2022 . title An elderly fall detection method based on federated learning and extreme learning machine (fed- ELM ) . journal IEEE Access volume 10 , pages 130816--130824

    2022

  67. [75]

    , author Xie, Y

    author Zhang, C. , author Xie, Y. , author Bai, H. , author Yu, B. , author Li, W. , author Gao, Y. , year 2021 . title A survey on federated learning . journal Knowledge-Based Systems volume 216 , pages 106775

    2021

  68. [76]

    , et al., year 2021

    author Zhao, Y. , et al., year 2021 . title Semi-supervised federated learning for activity recognition . journal ACM Trans. Intell. Syst. Technol volume 1

    2021

Pith tools

Reviewed August 6, 2026 · model on record in the stance chip above.