Pith. sign in

REVIEW 3 major objections 4 minor 29 references

Beyond Simulated Benchmarks: Evaluating Motion Representations for Fall Detection Under Real-World Data Scarcity

T0 review · 3 major / 4 minor · reviewed 2026-08-15 · deepseek-v4-flash

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

Pith's one-line read The paper claims that in-domain accuracy is a poor predictor of real-world robustness, and that anchoring motion representations to dataset-invariant physical quantities—specifically impact magnitude in units of g—is what lets a fall…

desk verdict A genuinely useful fall-detection benchmark with a credible simulation-to-reality warning, but the physically-grounded-transfer claim is selected on the test target and should not be treated as established. read the letter →

arxiv 2608.13197 v1 pith:5J6UQ43D submitted 2026-08-13 cs.LG

Timilehin B. Aderinola , Ilaria D'Ascanio , Luca Palmerini , Lorenzo Chiari , Jochen Klenk , Clemens Becker , Brian Caulfield , Georgiana Ifrim This is my paper · ORCID
classification cs.LG
keywords falldetectionwearableaccelerometermotionrepresentationreal-worlddatascarcitycross-datasettransfersymbolicaggregateapproximationimpactdescriptorsstreamingevaluation
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

Falls are hard to detect in practice because real-world falls are extremely rare: collecting 100 of them takes an estimated 100,000 days of monitoring, so most models train on simulated falls and fail when deployed. This paper systematically compares four motion-representation families—interval, kernel, symbolic, and foundation-model—under subject-wise cross-validation, controlled data scarcity, and simulated-to-real transfer on accelerometer data. It finds that the methods with the best simulated performance, kernel and foundation models, degrade most under domain shift, while FallLM, a lightweight symbolic representation that augments SAX tokens with physically grounded impact descriptors, loses only 9.5% of its F1 score in transfer and still detects falls after training on as few as two real examples. The central lesson is that representation choice, not model capacity, governs deployability, and that in-domain accuracy alone is a poor guide to real-world robustness.

What carries the argument

The load-bearing mechanism is FallLM's impact token: the peak acceleration magnitude in each window is discretized into three levels—low (<1.8g), medium (1.8–2.5g), high (>2.5g)—following biomechanical ranges, plus a relative impact token comparing the peak to the window's own magnitude distribution. Because these thresholds are expressed in gravitational units, their meaning is invariant across datasets and sensor placements; an ablation shows this coarse invariant encoding transfers far better than either no magnitude cue or a 12-bin magnitude fitted to the training data. These tokens are appended to a three-symbol SAX (Symbolic Aggregate Approximation) representation of the 3-second fall window, vectorized as TF-IDF n-grams, and classified by logistic regression, making every decision reducible to human-readable motion motifs such as 'b d c impact_med'.

What would settle it

Train FallLM on FallAllD and evaluate it on a second clinically verified real-world fall dataset recorded at a different sensor placement, such as wrist or thigh; if the cross-dataset F1 advantage over QUANT and Mantis shrinks or reverses, the claim that the g-anchored thresholds are dataset-invariant is falsified. A more immediate check is to re-run the Appendix A.1 ablation while holding out FARSEEING during token selection; if a token configuration chosen without peeking no longer beats SAX-only under transfer, the reported robustness is partly an artifact of selection.

Watch

Extended reading notes

Core claim

The paper's central claim is that robustness to the simulated-to-real gap in fall detection is determined not by model capacity but by the type of prior a motion representation encodes. Representations that learn dataset-specific signal statistics—MiniRocket's random kernels, Mantis's transformer embeddings, WEASEL's SFA dictionaries—reach F1 above 0.95 on simulated FallAllD falls but fall to 0.37–0.50 on real FARSEEING falls, with the biggest losses in recall. FallLM, which encodes each 3-second window as three SAX symbols plus an absolute impact token thresholded at 1.8g and 2.5g and a relative impact token, transfers with the smallest F1 drop (9.5%, from 0.74 simulated to 0.67 real) and is the only method whose precision improves across domains. The same token gives measurable detection (mean F1 ≈ 0.35) when trained on just two real falls, while the best abundant-data methods collapse under extreme scarcity. The paper is explicit that FallLM is a proof-of-principle rather than a deployable detector: it is imprecise in-domain (F1 0.64) and biased toward high-impact falls, so the conclusion is about what makes representations transfer, not that this particular model is ready for clinical use.

Load-bearing premise

The conclusions rest on FARSEEING being representative of real-world falls and on the 1.8g and 2.5g impact thresholds being truly dataset-invariant; moreover, the 3-level token was chosen via its transfer performance on FARSEEING, so the headline transfer result is not a fully independent test.

Editorial extensions

If this is right

  • Cross-dataset evaluation should become a standard reporting requirement: selecting a fall detector on simulated performance alone chooses among the least deployable options.
  • Under extreme data scarcity, representation choice matters more than model capacity: FallLM reaches mean F1 around 0.35 from two real training falls, while the other methods stay near or at zero.
  • Representations anchored to dataset-invariant physical quantities, such as impact magnitude in g, transfer more gracefully than representations that model dataset-specific signal statistics.
  • A deployable detector will need to combine impact-anchored robustness with sensitivity to low-impact falls, because FallLM's transfer advantage comes with in-domain precision loss and a high-impact bias.

Reading between the lines

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

  • We infer the principle generalizes beyond falls: any wearable event-detection task with scarce real labels could anchor features to physical units that are invariant across domains, such as g-force thresholds or joint-angle ranges.
  • We infer that a two-stage detector—keeping the invariant impact anchor for recall and adding a learned low-impact-sensitive stage—would address the two complementary failure modes the paper identifies.
  • We infer that the dataset-invariance claim for the 1.8g/2.5g boundaries is itself testable: re-running the transfer experiment on another real-world fall corpus with a different sensor placement (wrist, thigh, chest) would either support or bound the claim.
  • We infer that calibrating the impact thresholds on a small labelled real-world subset could further narrow the simulation-to-reality gap, since the paper deliberately studies only zero-shot transfer.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

3 major / 4 minor

Summary. The paper systematically evaluates four families of motion representations—interval-based (QUANT), kernel-based (MiniRocket), symbolic (WEASEL, MrSQM, FallLM), and foundation models (Mantis)—for wearable fall detection under real-world data scarcity. Using a streaming event-detection protocol with subject-wise splits, the authors compare methods on FallAllD (simulated falls) and FARSEEING (clinically verified real-world falls) via in-domain cross-validation, controlled data-scarcity experiments, and zero-shot cross-dataset transfer. The main empirical finding is that high simulated performance does not transfer to real-world data: methods with F1>0.95 on FallAllD drop to 0.37–0.62 on FARSEEING. The authors introduce FallLM, a lightweight SAX-based representation augmented with physically grounded impact tokens (absolute and relative), which achieves the smallest relative F1 drop under transfer (9.5%) and retains detection sensitivity with as few as two training falls. They argue that robustness tracks the type of representational prior, with dataset-invariant physical quantities transferring more gracefully than dataset-specific signal statistics.

Significance. If its central claims hold, this paper makes a valuable contribution to wearable fall detection by providing a principled, reproducible comparison of representation families under realistic deployment constraints. The streaming evaluation protocol, subject-wise splits, and cross-dataset transfer design are methodologically sound and address a real gap in the literature. The paper also ships code, supports token-level interpretability, and offers a concrete, falsifiable claim: coarse, physically grounded impact tokens transfer better than fine-grained, data-fitted magnitude tokens. The scarcity analysis is a useful addition. However, the load-bearing mechanism claim—that physical grounding, rather than coarse quantization or model simplicity, drives transfer robustness—is not independently tested because the winning 3-level token configuration was selected on the FARSEEING transfer metric itself.

major comments (3)
  1. [Section 3.4, Appendix A.1, Table A2] The headline transfer result is target-selected. Section 3.4 states that the 3-level impact token 'transferred best to FARSEEING (see Appendix A.1)', and Table A2 indeed compares the 3-level configuration against SAX-only and a 12-bin variant. Since the 3-level token was chosen after observing its FARSEEING transfer F1, the reported 9.5% degradation in Table 3 is not an independent test of the physical-prior hypothesis. This inflates the apparent advantage of FallLM and weakens the central claim that dataset-invariant physical quantities transfer more gracefully. A prespecified configuration, a nested model-selection procedure, or explicit reporting of selection bias would be needed to make this claim solid.
  2. [Appendix A.1, Table A2] The ablation confounds two factors: coarseness of the magnitude encoding and provenance of the thresholds. The 3-level token uses only three bins with literature thresholds (1.8g, 2.5g), while the 12-bin token uses twelve bins fitted to FallAllD's training distribution. A coarse token would be expected to survive distribution shift even without physical grounding, and a fine token fitted to the training population would be expected to lose recall on shifted data. Table A2 cannot distinguish 'dataset-invariant physical quantity' from 'coarse quantization is shift-robust'. The claim in Section 6 that 'what governs transfer is not model simplicity but the type of prior' requires ablations that vary one factor at a time, e.g., a 3-level token fitted to FallAllD versus the literature-threshold 3-level token, or a 12-bin token with literature-derived boundaries.
  3. [Section 6 / Limitations (final paragraph)] The paper's main empirical conclusion—'in-domain accuracy is a poor predictor of real-world robustness'—is based on a single simulated dataset (FallAllD) and a single real-world dataset (FARSEEING). The authors acknowledge this limitation in their final paragraph, but the claim is stated in general terms throughout the abstract and discussion. Since FARSEEING is the only clinically verified real-world corpus used, and the L5 subset comprises 150 falls, the generality of the simulation-to-reality gap remains uncertain. A second real-world dataset or a sensitivity analysis varying the choice of real-world test corpus would materially strengthen the claim. As written, the conclusion is an existence proof for this pair of datasets rather than a general law.
minor comments (4)
  1. [Several table captions] Displayed text contains 'T able' (Tables 1, 2, 3, 4 and others), which appears to be a LaTeX rendering artifact; please fix the formatting.
  2. [Section 5.2 / Figure 3] The scarcity curves report mean performance over five random subsets but no error bands or per-subset variability. Given the extreme 1% condition (two falls), reporting standard deviations or individual run values would help the reader assess the reliability of the claim that FallLM 'is the only method to achieve substantial detection'.
  3. [Appendix A.2, Table A4] The FARSEEING alphabet-size sweep shows n_bins=4 performing notably worse (F1=0.518) than all other settings, yet n_bins=5 is declared the default based on interpretability. The text says this sweep was 'not used to select n_bins', but the reader may wonder whether the paper's conclusions would change if n_bins were allowed to vary; a brief interpretive sentence on why n_bins=4's drop does not affect the representation-family comparison would improve clarity.
  4. [Section 3.4 / Table A1] The description of the relative impact token ('peak acceleration relative to the magnitude distribution within the window') is vague. Please specify the exact formula or algorithm used to compute the three relative levels, as this affects reproducibility.

Circularity Check

1 steps flagged · score 6.0 of 10

FallLM's headline cross-dataset advantage is target-selected: the 3-level impact token was chosen on the same FARSEEING transfer folds later reported as evidence, so the physical-anchoring claim is partly circular.

  1. fitted input called prediction [Section 3.4 and Appendix A.1 (Table A2); reported as result in Section 5.3 (Table 3) and Section 6]
    "Table A2 compares three impact-token configurations under transfer: no impact token (SAX only), the 3-level categorical token used in our final design (SAX + impact, 3-level), and a fine-grained 12-bin physical-magnitude token (SAX + impact, 12-bin) with bins learned from the training data. ... The 3-level categorical token avoids both failure modes, giving by far the best real-world performance (F1 = 0.667 ± 0.093) ..."

    As written, the ablation both identifies the 3-level token as 'our final design' and confirms it on the same FARSEEING transfer folds; Table 3 then reports that same configuration's transfer score (real F1 = 0.67; 9.5% drop) as the headline result, and Section 6 elevates it to the principle that 'representations anchored to dataset-invariant physical quantities transfer more gracefully.' Thus the reported prediction is the selection criterion, not an independent test.

full rationale

The evaluation as a whole is not circular: all methods are trained on FallAllD and tested on clinically verified FARSEEING labels, giving an external benchmark for the simulation-to-reality gap, and the overall ranking in Table 3 is a genuine comparison. The paper's central mechanistic claim, however, rests on one controlled comparison that is itself the selection step. Appendix A.1's Table A2 defines the 3-level token as 'used in our final design' after observing that it 'avoids both failure modes' on the FARSEEING transfer folds; Section 5.3 presents that same configuration's transfer score as the headline result, and Section 6 generalizes it to the principle that 'representations anchored to dataset-invariant physical quantities transfer more gracefully.' Because the token's bin count and threshold provenance are varied together, and because the winning configuration was selected on the very target used for evaluation, the physical-anchoring claim is not independently tested; it is in part a target-selected estimate. No load-bearing self-citation circularity was found: the cited prior work [1,2] supplies the streaming protocol and context, not the physical-prior premise, and the 1.8g/2.5g thresholds are attributed to external biomechanical work [10]. The absence of an independent validation of the chosen impact configuration is the main circular component, warranting a score of 6 rather than a higher score.

Assumptions & free parameters 4 free parameters · 5 assumptions · 0 invented entities

Free parameters are mostly fixed hyperparameters of FallLM. The most important is the impact token granularity, selected on the FARSEEING transfer target in Appendix A.1. Axioms are domain assumptions about dataset representativeness, the invariance of g-unit thresholds, and the streaming tolerance window. No new physical entities are proposed; the impact tokens are features, not entities with independent falsifiable handles.

free parameters (4)
  • SAX alphabet size (n_bins) = 5 (default; swept 3-10).
    Fixed FallLM hyperparameter. Appendix sweeps show F1 is stable across 3-10, so this choice is not driving the main result.
  • Impact token granularity = 3-level; chosen over none and 12-bin.
    Appendix A.1 reports that 3-level gives best FARSEEING transfer F1 (0.667), while SAX-only gives 0.418 and 12-bin gives 0.464. This selection on the evaluation target weakens the independence of the transfer claim.
  • PAA word size = 3.
    Maps the 3-second window to pre-impact, impact, and post-impact phases. Fixed a priori from the temporal structure of falls.
  • ADL retention threshold = 1.4 g.
    Negative samples below 1.4 g are dropped following Palmerini et al. [20]. This affects the empirical ADL distribution used in training and therefore the precision estimates.
assumptions (5)
  • domain assumption FARSEEING L5 subset (150 falls, 54 participants) is representative of clinically verified real-world falls with reliable labels.
    All in-domain and scarcity conclusions are drawn from this single dataset. If it is atypical, the findings may not transfer.
  • domain assumption Acceleration magnitude M removes sensor-orientation effects and makes waist (FallAllD) and L5 (FARSEEING) placements comparable.
    Used in preprocessing (Section 3.2) and required for the cross-dataset transfer experiment.
  • domain assumption The absolute impact thresholds (<1.8g, 1.8-2.5g, >2.5g), taken from [10], are dataset-invariant and remain meaningful across devices, placements, and populations.
    FallLM's physical anchoring and transfer advantage rest on these boundaries (Section 3.4).
  • domain assumption The streaming tolerance interval [f-3, f+20) seconds with IoU > 0 is a valid criterion for clinical fall detection.
    All precision, recall, and F1 numbers depend on this evaluation window (Section 4.2).
  • standard math SAX Gaussian breakpoints are a valid discretization for accelerometer magnitude PAA values.
    FallLM and MrSQM build on the standard SAX assumption of normality, which is an approximation for acceleration signals.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Beyond Simulated Benchmarks: Evaluating Motion Representations for Fall Detection Under Real-World Data Scarcity." pith.science (2026). https://pith.science/paper/5J6UQ43D

@misc{pith2026260813197,
  author       = {Pith},
  title        = {Pith review of: Beyond Simulated Benchmarks: Evaluating Motion Representations for Fall Detection Under Real-World Data Scarcity},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/5J6UQ43D}},
  note         = {Machine review of arXiv:2608.13197}
}
read the original abstract

Falls are a major health concern for older adults, and wearable sensors have been widely explored for detecting falls and enabling timely intervention. However, real-world falls are extremely rare: collecting 100 of them requires an estimated 100,000 days of monitoring, resulting in severely limited labelled data for training machine learning models. Consequently, many approaches rely on simulated datasets, often reporting high laboratory performance but limited real-world generalisation. We present a systematic evaluation of motion representations for wearable fall detection under real-world data scarcity. Using accelerometer signals, we compare interval-based, kernel-based, symbolic, and foundation model representations. As an interpretable baseline, we additionally investigate a lightweight symbolic representation that converts short motion segments into symbolic sentences augmented with physically-grounded impact descriptors. Experiments use FallAllD, a simulated falls dataset, and FARSEEING, a clinically verified real-world falls dataset. Through cross-validation, controlled data scarcity, and cross-dataset transfer, we examine how representation choices affect robustness under realistic deployment. Our results reveal that highly parameterised kernel and foundation models excel on simulated data but degrade severely under both data scarcity and domain shift. Although the interval-based representation achieves the strongest absolute real-world performance, augmenting a symbolic representation with physically-grounded impact descriptors yields the smallest degradation under domain shift and retains detection sensitivity under extreme scarcity, albeit at lower precision. These findings highlight the importance of evaluating beyond simulated benchmarks and show that representation choice is critical for deployable fall detection given the scarcity of real-world data.

Figures

Figures reproduced from arXiv: 2608.13197 by the authors.

Figure 1
Figure 1. A 30-second FARSEEING fall segment showing pre-fall activity, the annotated impact (dashed), and the post-fall lying period. Here, the detection window overlaps the ground-truth tolerance interval (shaded), counting as a true positive. This example is illustrative and not indicative of overall detection precision [PITH_FULL_IMAGE:figures/full_fig_p006_1.png] view at source ↗
Figure 2
Figure 2. Overview of the FallLM pipeline. signal at different temporal locations. These interval features are then used by an ensemble classifier (ExtraTrees Classifier). Foundation Model Representation. Mantis [8] is a transformer-based model pre-trained using contrastive learning to capture general temporal patterns across datasets. In our experiments, the pre-trained model is finetuned to generate embeddings for fall dete… view at source ↗
Figure 3
Figure 3. Performance under varying levels of real-world fall data availability (FARSEE￾ING). The x-axis represents the fraction of fall events used for training, while all ADL samples are retained. The shaded region highlights the extreme data scarcity regime (1–10% falls). Results are averaged across five randomly sampled subsets of fall events. training or adaptation. To enable a direct comparison between in-domain and cro… view at source ↗
Figures from the paper (1 more)
Figure 4
Figure 4. Figure 4: Impact-level distribution of the top-20 fall and ADL motifs on FallAllD and FARSEEING. Numbers are motif counts (out of 20). FARSEEING, where high impact is unambiguous, but which misses lower-impact real falls, capping its recall. 6 Discussion Our results point to a s…

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

29 extracted references · 25 canonical work pages

  1. [1]

    arXiv preprint arXiv:2509.11789 (2025)

    Aderinola, T.B., Palmerini, L., D’Ascanio, I., Chiari, L., Klenk, J., Becker, C., Caulfield, B., Ifrim, G.: Watch your step: A cost-sensitive framework for accelerometer-based fall detection in real-world streaming scenarios. arXiv preprint arXiv:2509.11789 (2025)

    work page arXiv 2025

  2. [2]

    In: International Workshop on Advanced Analytics and Learning on Temporal Data

    Aderinola, T.B., Palmerini, L., D’Ascanio, I., Chiari, L., Klenk, J., Becker, C., Caulfield, B., Ifrim, G.: Accurate and efficient real-world fall detection using time series techniques. In: International Workshop on Advanced Analytics and Learning on Temporal Data. pp. 52–79. Springer (2024)

    work page 2024

  3. [3]

    Biomedical Signal Processing and Control113, 108942 (2026)

    Bach, N.C., Van Chi, N., Cuong, D.D., Tuan, N.A., Kieu, T.Q.T., Phuong, N.T., Thien, N.D., Thao, L.Q.: A lightweight and efficient deep learning model for real- time fall detection on edge device. Biomedical Signal Processing and Control113, 108942 (2026)

    work page 2026

  4. [4]

    Clinical interventions in aging pp

    Camp, K., Murphy, S., Pate, B.: Integrating fall prevention strategies into ems services to reduce falls and associated healthcare costs for older adults. Clinical interventions in aging pp. 561–569 (2024)

    work page 2024

  5. [5]

    Data Mining and Knowledge Discovery34(5), 1454–1495 (2020)

    Dempster, A., Petitjean, F., Webb, G.I.: Rocket: exceptionally fast and accurate time series classification using random convolutional kernels. Data Mining and Knowledge Discovery34(5), 1454–1495 (2020)

    2020

  6. [6]

    In: Proceedings of the 27th ACM SIGKDD conference on knowledge discovery & data mining

    Dempster, A., Schmidt, D.F., Webb, G.I.: Minirocket: A very fast (almost) deter- ministic transform for time series classification. In: Proceedings of the 27th ACM SIGKDD conference on knowledge discovery & data mining. pp. 248–257 (2021)

    work page 2021

  7. [7]

    Data Mining and Knowledge Discovery pp

    Dempster, A., Schmidt, D.F., Webb, G.I.: Quant: A minimalist interval method for time series classification. Data Mining and Knowledge Discovery pp. 1–26 (2024)

    work page 2024

  8. [8]

    arXiv preprint arXiv:2502.15637 (2025)

    Feofanov, V., Wen, S., Alonso, M., Ilbert, R., Guo, H., Tiomoko, M., Pan, L., Zhang, J., Redko, I.: Mantis: Lightweight calibrated foundation model for user-friendly time series classification. arXiv preprint arXiv:2502.15637 (2025)

    arXiv 2025

Show all 29 references
  1. [9]

    Sensors23(13), 5845 (2023)

    Gil-Martín, M., López-Iniesta, J., Fernández-Martínez, F., San-Segundo, R.: Re- ducing the impact of sensor orientation variability in human activity recognition using a consistent reference system. Sensors23(13), 5845 (2023)

    2023

  2. [10]

    Signals2(1), 1–12 (2021)

    Huynh, Q.T., Tran, B.Q.: Time-frequency analysis of daily activities for fall detec- tion. Signals2(1), 1–12 (2021)

    2021

  3. [11]

    European review of aging and physical activity13, 1–7 (2016)

    Klenk, J., Schwickert, L., Palmerini, L., Mellone, S., Bourke, A., Ihlen, E.A., Kerse, N., Hauer, K., Pijnappels, M., Synofzik, M., et al.: The farseeing real-world fall repository: a large-scale collaborative database to collect and share sensor signals from real-world falls....

    2016

  4. [12]

    IEEE Sensors Journal 19(18), 8293–8302 (2019) 16 T.B

    Lee, J.S., Tseng, H.H.: Development of an enhanced threshold-based fall detection system using smartphones with built-in accelerometers. IEEE Sensors Journal 19(18), 8293–8302 (2019) 16 T.B. Aderinola et al

    2019

  5. [13]

    In: 2023 IEEE International Symposium on Medical Measurements and Applications (MeMeA)

    Liu, C.P., Li, J.H., Chu, E.P., Hsieh, C.Y., Liu, K.C., Chan, C.T., Tsao, Y.: Deep learning-based fall detection algorithm using ensemble model of coarse-fine cnn and gru networks. In: 2023 IEEE International Symposium on Medical Measurements and Applications (MeMeA). pp. 1–5....

    2023

  6. [14]

    Engineering Applications of Artificial Intelligence121, 105993 (2023)

    Liu, J., Li, X., Huang, S., Chao, R., Cao, Z., Wang, S., Wang, A., Liu, L.: A review of wearable sensors based fall-related recognition systems. Engineering Applications of Artificial Intelligence121, 105993 (2023)

    2023

  7. [15]

    arXiv preprint arXiv:2406.14231 (2024)

    Middlehurst,M.,Ismail-Fawaz,A.,Guillaume,A.,Holder,C.,Rubio,D.G.,Bulatova, G., Tsaprounis, L., Mentel, L., Walter, M., Schäfer, P., et al.: aeon: a python toolkit for learning from time series. arXiv preprint arXiv:2406.14231 (2024)

    2024 arXiv

  8. [16]

    IEEE Sensors Journal24(4), 4181–4198 (2024)

    Mohan, D., Al-Hamid, D.Z., Chong, P.H.J., Sudheera, K.L.K., Gutierrez, J., Chan, H.C., Li, H.: Artificial intelligence and iot in elderly fall prevention: A review. IEEE Sensors Journal24(4), 4181–4198 (2024)

    2024

  9. [17]

    Vietnam Journal of Computer Science pp

    Nguyen, D.A., Pham, C., Argent, R., Caulfield, B., Le-Khac, N.A.: Model and empirical study on multi-tasking learning for human fall detection. Vietnam Journal of Computer Science pp. 1–14 (2024)

    2024

  10. [18]

    In: International Workshop on Advanced Analytics and Learning on Temporal Data

    Nguyen, T.L., Ifrim, G.: Fast time series classification with random symbolic subsequences. In: International Workshop on Advanced Analytics and Learning on Temporal Data. pp. 50–65. Springer (2022)

    2022

  11. [19]

    Digital Health11, 20552076251386698 (2025)

    Owusu, E., Acquah, I., Asare, M.A., Yeboah, B.A.: Litefallnet: A lightweight deep learning model for efficient real-time fall detection. Digital Health11, 20552076251386698 (2025)

    2025

  12. [20]

    Sensors20(22), 6479 (2020)

    Palmerini, L., Klenk, J., Becker, C., Chiari, L.: Accelerometer-based fall detection using machine learning: Training and testing on real-world falls. Sensors20(22), 6479 (2020)

    2020

  13. [21]

    Computational intelligence37(2), 951–974 (2021)

    Rastogi, S., Singh, J.: A systematic review on machine learning for fall detection system. Computational intelligence37(2), 951–974 (2021)

    2021

  14. [22]

    IEEE Sensors Journal21(2), 1849–1858 (2020)

    Saleh, M., Abbas, M., Le Jeannes, R.B.: Fallalld: An open dataset of human falls and activities of daily living for classical and deep learning applications. IEEE Sensors Journal21(2), 1849–1858 (2020)

    2020

  15. [23]

    Machine Learning 112(12), 4763–4788 (2023)

    Schäfer, P., Leser, U.: Weasel 2.0: a random dilated dictionary transform for fast, accurate and memory constrained time series classification. Machine Learning 112(12), 4763–4788 (2023)

    2023

  16. [24]

    Sensors17(1), 198 (2017)

    Sucerquia, A., López, J.D., Vargas-Bonilla, J.F.: Sisfall: A fall and movement dataset. Sensors17(1), 198 (2017)

    2017

  17. [25]

    Data Mining and Knowledge Discovery36(5), 1623–1646 (2022)

    Tan, C.W., Dempster, A., Bergmeir, C., Webb, G.I.: Multirocket: multiple pooling operators and transformations for fast and effective time series classification: Cw tan. Data Mining and Knowledge Discovery36(5), 1623–1646 (2022)

    2022

  18. [26]

    In: 2025 4th International Conference on Sentiment Analysis and Deep Learning (ICSADL)

    Vignesh, P.G., Kumar, N., et al.: Lidar-based elderly fall detection system for indoor environments using neural network algorithms. In: 2025 4th International Conference on Sentiment Analysis and Deep Learning (ICSADL). pp. 1564–1568. IEEE (2025)

    2025

  19. [27]

    World Health Organization: Step safely: strategies for preventing and managing falls across the life-course (2021)

    2021

  20. [28]

    BMJ Health & Care Informatics 32(1), e101439 (2025)

    Zafar, R.O., Zafar, F.: Real-time activity and fall detection using transformer-based deep learning models for elderly care applications. BMJ Health & Care Informatics 32(1), e101439 (2025)

    2025

  21. [29]

    Zhang, J., Li, Z., Liu, Y., Li, J., Qiu, H., Li, M., Hou, G., Zhou, Z.: An effective deep learning framework for fall detection: model development and study design. Journal of medical internet research26, e56750 (2024) Motion Representations for Fall Detection Under Real-World...

    2024

Pith tools

Reviewed August 15, 2026 · model on record in the stance chip above.