pith. sign in

arxiv: 2604.09693 · v1 · submitted 2026-04-06 · 💻 cs.CV · cs.AI

TaFall: Balance-Informed Fall Detection via Passive Thermal Sensing

Chengxiao Li , Xie Zhang , Wei Zhu , Yan Jiang , Chenshu Wu This is my paper

Pith reviewed 2026-05-10 19:06 UTC · model grok-4.3

classification 💻 cs.CV cs.AI
keywords fall detectionthermal sensingbalance dynamicspose estimationprivacy-preserving monitoringelder caremotion analysishome deployment
0
0 comments X

The pith

Falls are detected by estimating biomechanical balance degradation from low-resolution thermal images.

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

The paper introduces TaFall, a system that treats falls as a process of balance loss rather than sudden motion alone. It reconstructs a person's pose from thermal array data by fusing appearance and motion cues, then applies physically grounded learning to estimate balance dynamics and flag when balance degrades into a fall. Special pretraining helps the model handle the limited detail in low-resolution heat maps. Large-scale tests with over 3,000 falls across many participants and real multi-day home deployments show very high detection rates paired with extremely low false alarms, even under moisture and thermal noise. This approach matters because most falls occur indoors where privacy constraints rule out cameras and wearables often prove impractical.

Core claim

TaFall shows that modeling a fall as progressive balance degradation, estimated via pose-driven biomechanical dynamics from passive thermal arrays, enables reliable detection. The system combines an appearance-motion fusion model for pose reconstruction, physically grounded balance-aware learning, and pose-bridged pretraining to overcome low image resolution. On a dataset of over 3,000 fall instances from 35 participants in varied indoor settings, it reaches 98.26% detection with 0.65% false alarms; 27-day deployments in four homes yield an ultra-low false alarm rate of 0.00126%, with additional tests confirming robustness to bathroom moisture and thermal interference.

What carries the argument

Appearance-motion fusion model for pose reconstruction paired with physically grounded balance-aware learning to estimate pose-driven biomechanical balance dynamics from thermal array maps.

If this is right

  • Continuous indoor monitoring becomes feasible without cameras or body-worn devices while respecting privacy.
  • The same thermal data can support detection across diverse room layouts and participant body types.
  • Extended home use maintains low false alarms over weeks rather than just lab sessions.
  • The method tolerates real bathroom conditions involving moisture and temperature changes.

Where Pith is reading between the lines

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

  • The balance-degradation framing could extend to spotting unsteady gait or near-falls before a complete fall occurs.
  • Similar low-resolution thermal pipelines might apply to other indoor safety tasks such as detecting prolonged immobility.
  • Long-term collection of balance metrics could reveal gradual health changes linked to fall risk.
  • The fusion and pretraining techniques may transfer to other inexpensive sensors that produce sparse spatial data.

Load-bearing premise

Balance degradation can be judged accurately enough from fuzzy thermal heat maps of a person's pose to distinguish real falls even when environments add moisture or thermal noise.

What would settle it

A controlled test in homes with high steam, unusual furniture blocking views, or participants using atypical movements that shows either many missed falls or a sharp rise in false alarms.

Figures

Figures reproduced from arXiv: 2604.09693 by Chengxiao Li, Chenshu Wu, Wei Zhu, Xie Zhang, Yan Jiang.

Figure 1
Figure 1. Figure 1: Illustration of thermal array temperature maps and the corresponding body pose–based balance representation [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Thermal array sensing characteristics and biomechanical context. [PITH_FULL_IMAGE:figures/full_fig_p005_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Overview of the TaFall framework. TaFall consists of two complementary phases. (a) Balance-Informed Fall Detection: an Appearance–Motion Fusion module estimates a robust 2.5D human pose sequence from thermal array temperature maps by jointly leveraging spatial appearance cues and motion-blur–induced dynamics. The resulting pose sequence is then processed by a Balance-Aware Pose Network, which learns a Phys… view at source ↗
Figure 4
Figure 4. Figure 4: The structure of the Spatial Branch. It first predicts the Center Heatmap and the Human Scale Map from the thermal [PITH_FULL_IMAGE:figures/full_fig_p007_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: The structure of Motion Blur Branch. It first generates the [PITH_FULL_IMAGE:figures/full_fig_p008_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: The illustration of the Physically Grounded Balance Representation and the Balance-Aware Pose Network. (a) [PITH_FULL_IMAGE:figures/full_fig_p009_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: Illustration of the Pose-Bridged OOV Enhancement strategy. A motion-capture dataset provides 3D skeletal trajectories [PITH_FULL_IMAGE:figures/full_fig_p011_7.png] view at source ↗
Figure 8
Figure 8. Figure 8: Experimental setup. Fig. (a) illustrates the three representative indoor environments used for fall data collection. [PITH_FULL_IMAGE:figures/full_fig_p013_8.png] view at source ↗
Figure 11
Figure 11. Figure 11: Cross user performance. 5.2 Performance of TaFall Overall Performance: We evaluate the overall performance of TaFall and compare it with representative baselines. As shown in [PITH_FULL_IMAGE:figures/full_fig_p015_11.png] view at source ↗
Figure 13
Figure 13. Figure 13: Results across direction. 15x20 31x40 62x80 0 20 40 60 80 100 DR (%) DR FAR 0 2 4 6 8 10 FAR (%) 15x20 31x40 62x80 [PITH_FULL_IMAGE:figures/full_fig_p016_13.png] view at source ↗
Figure 15
Figure 15. Figure 15: Occlusion of different objects and the impact of object occlusion [PITH_FULL_IMAGE:figures/full_fig_p016_15.png] view at source ↗
Figure 18
Figure 18. Figure 18: Performance of fall-like activities. body is partially obstructed by surrounding objects, especially during the later stages of a fall when the subject approaches the ground, TaFall may fail to reliably detect the human body, as a large portion of the body is occluded in the thermal observations. Despite this limitation, the proposed system remains capable of accurately inferring balance states from incom… view at source ↗
Figure 19
Figure 19. Figure 19: Long-term false alarm case study in real environment. Fig. (a) shows five types of false alarms(FA), where the red [PITH_FULL_IMAGE:figures/full_fig_p018_19.png] view at source ↗
Figure 20
Figure 20. Figure 20: Bathroom case study. (a) Examples of collected daily activities and falls in bathroom environment. (b) False alarm [PITH_FULL_IMAGE:figures/full_fig_p019_20.png] view at source ↗
read the original abstract

Falls are a major cause of injury and mortality among older adults, yet most incidents occur in private indoor environments where monitoring must balance effectiveness with privacy. Existing privacy-preserving fall detection approaches, particularly those based on radio frequency sensing, often rely on coarse motion cues, which limits reliability in real-world deployments. We introduce TaFall, a balance-informed fall detection system based on low-cost, privacy-preserving thermal array sensing. The key insight is that TaFall models a fall as a process of balance degradation and detects falls by estimating pose-driven biomechanical balance dynamics. To enable this capability from low-resolution thermal array maps, we propose (i) an appearance-motion fusion model for robust pose reconstruction, (ii) physically grounded balance-aware learning, and (iii) pose-bridged pretraining to improve robustness. TaFall achieves a detection rate of 98.26% with a false alarm rate of 0.65% on our dataset with over 3,000 fall instances from 35 participants across diverse indoor environments. In 27 day deployments across four homes, TaFall attains an ultra-low false alarm rate of 0.00126% and a pilot bathroom study confirms robustness under moisture and thermal interference. Together, these results establish TaFall as a reliable and privacy-preserving approach to fall detection in everyday living environments.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

2 major / 2 minor

Summary. The paper introduces TaFall, a privacy-preserving fall detection system using low-cost thermal array sensors. It models falls as a process of balance degradation and estimates pose-driven biomechanical balance dynamics via an appearance-motion fusion model for pose reconstruction, physically grounded balance-aware learning, and pose-bridged pretraining. The system reports 98.26% detection rate and 0.65% false alarm rate on a dataset of >3000 falls from 35 participants across indoor environments, plus ultra-low false alarms (0.00126%) in 27-day deployments across four homes and robustness in a bathroom pilot study.

Significance. If the intermediate pose reconstruction and balance dynamics estimation prove accurate, TaFall could provide a meaningful advance in real-world, privacy-preserving fall detection for older adults, with strong empirical grounding in large-scale data collection and multi-home deployments that go beyond lab-only results.

major comments (2)
  1. [Evaluation / Experiments] The central claim requires that pose-driven biomechanical balance dynamics are reliably recovered from low-resolution thermal arrays. However, the evaluation provides no quantitative pose reconstruction accuracy metrics (e.g., MPJPE, PCK, or correlation of derived balance features against RGB/D or motion-capture ground truth) to validate the appearance-motion fusion model. Without these, the 98.26% detection rate and deployment results rest on an untested intermediate representation.
  2. [Methods / Balance-aware learning] The balance-aware learning component is described as 'physically grounded,' yet the manuscript does not detail the exact biomechanical features extracted from the estimated poses, the loss formulation, or any ablation isolating the contribution of the balance model versus appearance-motion fusion alone. This makes it difficult to assess whether the performance gains are attributable to the claimed balance modeling.
minor comments (2)
  1. [Abstract] The abstract and introduction would benefit from explicitly stating the thermal array resolution (e.g., 8x8, 16x16, or 32x32) used in the hardware to better contextualize the low-resolution challenge.
  2. [Results / Deployments] Table or figure captions for the deployment results should include the exact number of false alarms observed and total monitoring hours to allow direct verification of the reported 0.00126% false alarm rate.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the thoughtful and constructive review. The comments highlight important aspects of validation and methodological clarity that we will address in the revision. We respond to each major comment below.

read point-by-point responses

  1. Referee: [Evaluation / Experiments] The central claim requires that pose-driven biomechanical balance dynamics are reliably recovered from low-resolution thermal arrays. However, the evaluation provides no quantitative pose reconstruction accuracy metrics (e.g., MPJPE, PCK, or correlation of derived balance features against RGB/D or motion-capture ground truth) to validate the appearance-motion fusion model. Without these, the 98.26% detection rate and deployment results rest on an untested intermediate representation.

    Authors: We agree that quantitative validation of the intermediate pose reconstruction would strengthen the paper. The current manuscript prioritizes end-to-end fall detection performance and real-home deployment results, as synchronized high-resolution ground truth (RGB or mocap) is difficult to obtain at scale in the same home environments used for thermal data collection. In the revised version, we will add a dedicated evaluation subsection reporting MPJPE and PCK on a held-out subset where synchronized RGB ground truth was collected, along with correlation analysis of the derived balance features (center-of-mass trajectory and stability margin). This will directly quantify the accuracy of the appearance-motion fusion model. revision: yes

  2. Referee: [Methods / Balance-aware learning] The balance-aware learning component is described as 'physically grounded,' yet the manuscript does not detail the exact biomechanical features extracted from the estimated poses, the loss formulation, or any ablation isolating the contribution of the balance model versus appearance-motion fusion alone. This makes it difficult to assess whether the performance gains are attributable to the claimed balance modeling.

    Authors: We acknowledge that the current manuscript provides only a high-level description of the balance-aware learning component. In the revision, we will expand the Methods section with: (1) the precise biomechanical features computed from the reconstructed poses (center-of-mass height and velocity, base-of-support area, and extrapolated center-of-mass stability margin); (2) the full loss formulation combining the balance regression term with the pose reconstruction loss; and (3) an ablation study that isolates the contribution of the balance-aware loss by comparing the full model against a variant trained with appearance-motion fusion alone. These additions will allow readers to assess the specific role of the physically grounded balance modeling. revision: yes

Circularity Check

0 steps flagged

No circularity: performance rests on independent empirical testing and deployments

full rationale

The paper's core claims derive from training and evaluating an appearance-motion fusion model plus balance-aware learning on a held-out dataset of >3000 fall instances from 35 participants, followed by separate 27-day home deployments and a bathroom pilot. No equations, parameters, or self-citations are shown to reduce the reported detection/false-alarm rates to quantities defined by the inputs themselves; the balance-dynamics estimation is an intermediate learned representation whose accuracy is assessed end-to-end via external ground-truth labels rather than by construction. The derivation chain therefore remains self-contained against external benchmarks.

Axiom & Free-Parameter Ledger

1 free parameters · 2 axioms · 0 invented entities

The approach rests on domain assumptions about balance modeling from poses and the informativeness of thermal data, plus standard machine learning free parameters in the neural models.

free parameters (1)
  • Hyperparameters and weights in appearance-motion fusion, balance-aware, and pretraining models
    Numerous parameters are optimized on the fall dataset to achieve the reported performance.
axioms (2)
  • domain assumption Falls can be modeled as processes of balance degradation detectable through pose-driven biomechanical dynamics
    This is the key insight that enables the entire detection approach.
  • domain assumption Low-resolution thermal array maps contain sufficient information for robust pose reconstruction via appearance-motion fusion
    Required to make the core capability work from the chosen sensor type.

pith-pipeline@v0.9.0 · 5539 in / 1434 out tokens · 93053 ms · 2026-05-10T19:06:59.662513+00:00 · methodology

discussion (0)

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

Reference graph

Works this paper leans on

71 extracted references · 71 canonical work pages

  1. [1]

    Syed M Adnan, Aun Irtaza, Sumair Aziz, M Obaid Ullah, Ali Javed, and Muhammad Tariq Mahmood. 2018. Fall detection through acoustic local ternary patterns.Applied Acoustics140 (2018), 296–300

    work page 2018

  2. [2]

    Jindrich Adolf, Martin Macas, Lenka Lhotska, and Jaromir Dolezal. 2018. Deep neural network based body posture recognitions and fall detection from low resolution infrared array sensor. In2018 IEEE International Conference on Bioinformatics and Biomedicine (BIBM). IEEE, 2394–2399

    work page 2018

  3. [3]

    Fabio Bagala, Clemens Becker, Angelo Cappello, Lorenzo Chiari, Kamiar Aminian, Jeffrey M Hausdorff, Wiebren Zijlstra, and Jochen Klenk. 2012. Evaluation of accelerometer-based fall detection algorithms on real-world falls.PloS one7, 5 (2012), e37062

    work page 2012

  4. [4]

    Bobick and James W

    Aaron F. Bobick and James W. Davis. 2002. The recognition of human movement using temporal templates.IEEE Transactions on pattern analysis and machine intelligence23, 3 (2002), 257–267

    work page 2002

  5. [5]

    Alan K Bourke, Jochen Klenk, Lars Schwickert, Kamiar Aminian, Espen AF Ihlen, Sabato Mellone, Jorunn L Helbostad, Lorenzo Chiari, and Clemens Becker. 2016. Fall detection algorithms for real-world falls harvested from lumbar sensors in the elderly population: A machine learning approach. In2016 38th annual international conference of the ieee engineering ...

    work page 2016

  6. [6]

    Wei-Han Chen and Hsi-Pin Ma. 2015. A fall detection system based on infrared array sensors with tracking capability for the elderly at home. In2015 17th International Conference on E-health Networking, Application & Services (HealthCom). IEEE, 428–434

    work page 2015

  7. [7]

    Yangsen Chen, Rongxi Du, Kaitao Luo, and Yuheng Xiao. 2021. Fall detection system based on real-time pose estimation and SVM. In 2021 IEEE 2nd international conference on big data, artificial intelligence and internet of things engineering (ICBAIE). IEEE, 990–993

    work page 2021

  8. [8]

    Man Cheung Chung, Kevin J McKee, Chris Austin, Helen Barkby, Hayley Brown, Sophie Cash, Jenny Ellingford, Liz Hanger, and Tara Pais. 2009. Posttraumatic stress disorder in older people after a fall.International Journal of Geriatric Psychiatry: A journal of the psychiatry of late life and allied sciences24, 9 (2009), 955–964

    work page 2009

  9. [9]

    Kaiwen Duan, Song Bai, Lingxi Xie, Honggang Qi, Qingming Huang, and Qi Tian. 2019. Centernet: Keypoint triplets for object detection. InProceedings of the IEEE/CVF international conference on computer vision. 6569–6578

    work page 2019

  10. [10]

    Yugma PN Fernando, Kasun DB Gunasekara, Kumary P Sirikumara, Upeksha E Galappaththi, Thusithanjana Thilakarathna, and Dharshana Kasthurirathna. 2021. Computer vision based privacy protected fall detection and behavior monitoring system for the care of the elderly. In2021 26th IEEE International Conference on Emerging Technologies and Factory Automation (E...

    work page 2021

  11. [11]

    Akira Hayashida, Vasily Moshnyaga, and Koji Hashimoto. 2017. The use of thermal ir array sensor for indoor fall detection. In2017 IEEE international conference on systems, man, and cybernetics (SMC). IEEE, 594–599

    work page 2017

  12. [12]

    At L Hof, MGJ Gazendam, and WE Sinke. 2005. The condition for dynamic stability.Journal of biomechanics38, 1 (2005), 1–8

    work page 2005

  13. [13]

    Shuting Hu, Siyang Cao, Nima Toosizadeh, Jennifer Barton, Melvin G Hector, and Mindy J Fain. 2024. Radar-based fall detection: A survey [survey].IEEE robotics & automation magazine31, 3 (2024), 170–185

    work page 2024

  14. [14]

    Xinyao Hu and Xingda Qu. 2016. Pre-impact fall detection.Biomedical engineering online15 (2016), 1–16

    work page 2016

  15. [15]

    Yuqian Hu, Feng Zhang, Chenshu Wu, Beibei Wang, and KJ Ray Liu. 2021. DeFall: Environment-independent passive fall detection using WiFi.IEEE Internet of Things Journal9, 11 (2021), 8515–8530

    work page 2021

  16. [16]

    Anitha Rani Inturi, VM Manikandan, and Vignesh Garrapally. 2023. A novel vision-based fall detection scheme using keypoints of human skeleton with long short-term memory network.Arabian Journal for Science and Engineering48, 2 (2023), 1143–1155

    work page 2023

  17. [17]

    Sijie Ji, Yaxiong Xie, and Mo Li. 2022. SiFall: Practical online fall detection with RF sensing. InProceedings of the 20th ACM Conference on Embedded Networked Sensor Systems. 563–577. 22 TaFall Arxiv, 2026

    work page 2022

  18. [18]

    Yan Jiang, Tianyi Gong, Lingfeng He, Shicheng Yan, Xiaoping Wu, and Jianyang Liu. 2024. Fall detection on embedded platform using infrared array sensor for healthcare applications.Neural Computing and Applications36, 9 (2024), 5093–5108

    work page 2024

  19. [19]

    Muhammad Salman Khan, Miao Yu, Pengming Feng, Liang Wang, and Jonathon Chambers. 2015. An unsupervised acoustic fall detection system using source separation for sound interference suppression.Signal processing110 (2015), 199–210

    work page 2015

  20. [20]

    Lachaux Killian, Maitre Julien, Bouchard Kevin, Lussier Maxime, Bottari Carolina, Couture Mélanie, Bier Nathalie, Giroux Sylvain, and Gaboury Sebastien. 2021. Fall prevention and detection in smart homes using monocular cameras and an interactive social robot. In Proceedings of the Conference on Information Technology for Social Good. 7–12

    work page 2021

  21. [21]

    Colin Lea, Rene Vidal, Austin Reiter, and Gregory D Hager. 2016. Temporal convolutional networks: A unified approach to action segmentation. InEuropean conference on computer vision. Springer, 47–54

    work page 2016

  22. [22]

    Chengxiao Li, Xie Zhang, and Chenshu Wu. 2025. Facial Expression Recognition with DToF Sensing. InICASSP 2025-2025 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP). IEEE, 1–5

    work page 2025

  23. [23]

    Haobo Li, Aman Shrestha, Francesco Fioranelli, Julien Le Kernec, Hadi Heidari, Matteo Pepa, Enea Cippitelli, Ennio Gambi, and Susanna Spinsante. 2017. Multisensor data fusion for human activities classification and fall detection. In2017 IEEE sensors. IEEE, 1–3

    work page 2017

  24. [24]

    Shuwei Li, Changhai Man, Ao Shen, Ziyi Guan, Wei Mao, Shaobo Luo, Rumin Zhang, and Hao Yu. 2022. A fall detection network by 2d/3d spatio-temporal joint models with tensor compression on edge.ACM Transactions on Embedded Computing Systems21, 6 (2022), 1–19

    work page 2022

  25. [25]

    Jie Lian, Xu Yuan, Ming Li, and Nian-Feng Tzeng. 2021. Fall detection via inaudible acoustic sensing.Proceedings of the ACM on Interactive, Mobile, Wearable and Ubiquitous Technologies5, 3 (2021), 1–21

    work page 2021

  26. [26]

    Liang Liu, Mihail Popescu, KC Ho, Marjorie Skubic, and Marilyn Rantz. 2012. Doppler radar sensor positioning in a fall detection system. In2012 annual international conference of the IEEE engineering in medicine and biology society. IEEE, 256–259

    work page 2012

  27. [27]

    Liang Liu, Mihail Popescu, Marjorie Skubic, Marilyn Rantz, Tarik Yardibi, and Paul Cuddihy. 2011. Automatic fall detection based on Doppler radar motion signature. In2011 5th International Conference on Pervasive Computing Technologies for Healthcare (PervasiveHealth) and Workshops. IEEE, 222–225

    work page 2011

  28. [28]

    Xuan Liu, Xie Zhang, and Chenshu Wu. 2025. Privacy-Preserving Non-Contact Sleep Monitoring via Multimodal Thermal-Depth Sensing. InProceedings of the 2025 ACM International Workshop on Thermal Sensing and Computing. 1–6

    work page 2025

  29. [29]

    Zhixin Liu, Ming Yang, Yazhou Yuan, and Kit Yan Chan. 2020. Fall detection and personnel tracking system using infrared array sensors. IEEE Sensors Journal20, 16 (2020), 9558–9566

    work page 2020

  30. [30]

    Lord, Catherine Sherrington, Hylton B

    Stephen R. Lord, Catherine Sherrington, Hylton B. Menz, and Jacqueline C. T. Close. 2007.Falls in Older People: Risk Factors and Strategies for Prevention(2 ed.). Cambridge University Press

    work page 2007

  31. [31]

    Chengzhen Meng, Chenming He, Dequan Wang, Yuxuan Xiao, Lingyu Wang, Xiaoran Fan, Lu Zhang, and Yanyong Zhang. 2025. Gr-fall: A fall detection system with gait recognition for indoor environments using siso mmwave radar.Proceedings of the ACM on Interactive, Mobile, Wearable and Ubiquitous Technologies9, 3 (2025), 1–26

    work page 2025

  32. [32]

    Briana L Moreland, Ramakrishna Kakara, Yara K Haddad, Iju Shakya, and Gwen Bergen. 2021. A descriptive analysis of location of older adult falls that resulted in emergency department visits in the United States, 2015.American journal of lifestyle medicine15, 6 (2021), 590–597

    work page 2021

  33. [33]

    Tamara Mujirishvili, Caterina Maidhof, Francisco Florez-Revuelta, Martina Ziefle, Miguel Richart-Martinez, and Julio Cabrero-García

    work page

  34. [34]

    Acceptance and privacy perceptions toward video-based active and assisted living technologies: scoping review.Journal of Medical Internet Research25 (2023), e45297

    work page 2023

  35. [35]

    Ahmed Nait Aicha, Gwenn Englebienne, Kimberley S Van Schooten, Mirjam Pijnappels, and Ben Kröse. 2018. Deep learning to predict falls in older adults based on daily-life trunk accelerometry.Sensors18, 5 (2018), 1654

    work page 2018

  36. [36]

    Abdallah Naser, Ahmad Lotfi, and Junpei Zhong. 2022. Multiple thermal sensor array fusion toward enabling privacy-preserving human monitoring applications.IEEE Internet of Things Journal9, 17 (2022), 16677–16688

    work page 2022

  37. [37]

    Nishat Tasnim Newaz and Eisuke Hanada. 2025. An Approach to Fall Detection Using Statistical Distributions of Thermal Signatures Obtained by a Stand-Alone Low-Resolution IR Array Sensor Device.Sensors (Basel, Switzerland)25, 2 (2025), 504

    work page 2025

  38. [38]

    Yoosuf Nizam and M Mahadi Abdul Jamil. 2020. Classification of daily life activities for human fall detection: a systematic review of the techniques and approaches.Challenges and Trends in Multimodal Fall Detection for Healthcare(2020), 137–179

    work page 2020

  39. [39]

    World Health Organization et al. 2008. WHO global report on falls prevention in older age World Health Organization

    work page 2008

  40. [40]

    Oswald-Tranta

    B. Oswald-Tranta. 2017. Motion Deblurring of Infrared Images. InProceedings IRS22017. AMA Conferences, Nürnberg, Germany, 783–787. https://doi.org/10.5162/irs2017/i3.1

    work page doi:10.5162/irs2017/i3.1 2017

  41. [41]

    Sameera Palipana, David Rojas, Piyush Agrawal, and Dirk Pesch. 2018. FallDeFi: Ubiquitous fall detection using commodity Wi-Fi devices.Proceedings of the ACM on Interactive, Mobile, Wearable and Ubiquitous Technologies1, 4 (2018), 1–25

    work page 2018

  42. [42]

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

    work page 2020

  43. [43]

    Adam Paszke, Sam Gross, Francisco Massa, Adam Lerer, James Bradbury, Gregory Chanan, Trevor Killeen, Zeming Lin, Natalia Gimelshein, Luca Antiga, et al . 2019. Pytorch: An imperative style, high-performance deep learning library.Advances in neural 23 Arxiv, 2026 TaFall information processing systems32 (2019)

    work page 2019

  44. [44]

    Aryan Nikul Patel, Ramalingam Murugan, Praveen Kumar Reddy Maddikunta, Gokul Yenduri, Rutvij H Jhaveri, Yaodong Zhu, and Thippa Reddy Gadekallu. 2024. AI-powered trustable and explainable fall detection system using transfer learning.Image and Vision Computing149 (2024), 105164

    work page 2024

  45. [45]

    Pietro Picerno, Marco Iosa, Clive D’Souza, Maria Grazia Benedetti, Stefano Paolucci, and Giovanni Morone. 2021. Wearable inertial sensors for human movement analysis: a five-year update.Expert review of medical devices18, sup1 (2021), 79–94

    work page 2021

  46. [46]

    Mihail Popescu, Yun Li, Marjorie Skubic, and Marilyn Rantz. 2008. An acoustic fall detector system that uses sound height information to reduce the false alarm rate. In2008 30th Annual International Conference of the IEEE Engineering in Medicine and Biology Society. IEEE, 4628–4631

    work page 2008

  47. [47]

    VM Pranavan, Maunika Shekar, Ruzelita Ngadiran, Sindhu Ravindran, et al. 2023. Fall detection system for monitoring elderly people using YOLOv7-pose detection model. In2023 International Conference on Computer, Electronics & Electrical Engineering & their Applications (IC2E3). IEEE, 1–6

    work page 2023

  48. [48]

    Ali Raza, Azam Mehmood Qadri, Iqra Akhtar, Nagwan Abdel Samee, and Maali Alabdulhafith. 2023. LogRF: An approach to human pose estimation using skeleton landmarks for physiotherapy fitness exercise correction.IEEE Access(2023)

    work page 2023

  49. [49]

    Ariyamehr Mohsen Rezaei, Michael C Stevens, Ahmadreza Argha, Alessandro Mascheroni, Alessandro Puiatti, and Nigel H Lovell. 2021. An unobtrusive fall detection system using low resolution thermal sensors and convolutional neural networks. In2021 43rd Annual International Conference of the IEEE Engineering in Medicine & Biology Society (EMBC). IEEE, 6949–6952

    work page 2021

  50. [50]

    Beddiar Djamila Romaissa, Oussalah Mourad, Nini Brahim, and Bounab Yazid. 2020. Fall detection using body geometry in video sequences. In2020 Tenth International Conference on Image Processing Theory, Tools and Applications (IPTA). IEEE, 1–5

    work page 2020

  51. [51]

    Hamidreza Sadreazami, Miodrag Bolic, and Sreeraman Rajan. 2019. Fall detection using standoff radar-based sensing and deep convolutional neural network.IEEE Transactions on Circuits and Systems II: Express Briefs67, 1 (2019), 197–201

    work page 2019

  52. [52]

    Rajkumar Saini, Pradeep Kumar, Barjinder Kaur, Partha Pratim Roy, Debi Prosad Dogra, and KC Santosh. 2019. Kinect sensor-based interaction monitoring system using the BLSTM neural network in healthcare.International Journal of Machine Learning and Cybernetics 10 (2019), 2529–2540

    work page 2019

  53. [53]

    Paul Scuffham, Stephen Chaplin, and Rosa Legood. 2003. Incidence and costs of unintentional falls in older people in the United Kingdom.Journal of Epidemiology & Community Health57, 9 (2003), 740–744

    work page 2003

  54. [54]

    A Shrestha, J Le Kernec, F Fioranelli, E Cippitelli, E Gambi, and S Spinsante. 2017. Feature diversity for fall detection and human indoor activities classification using radar systems. InInternational Conference on Radar Systems (Radar 2017). IET, 1–6

    work page 2017

  55. [55]

    Wei-Chih Su, Xuan-Xin Wu, Tzyy-Sheng Horng, and Mu-Cyun Tang. 2022. Hybrid continuous-wave and self-injection-locking monopulse radar for posture and fall detection.IEEE Transactions on Microwave Theory and Techniques70, 3 (2022), 1686–1695

    work page 2022

  56. [56]

    Dongsheng uan, Xie Zhang, Weiying Hou, Sheng Lyu, Yuemin Yu, Luca Jiang-Tao Yu, Chengxiao Li, and Chenshu Wu. 2025. OctoNet: A Large-Scale Multi-Modal Dataset for Human Activity Understanding Grounded in Motion-Captured 3D Pose Labels. The Thirty-ninth Annual Conference on Neural Information Processing Systems Datasets and Benchmarks Track

    work page 2025

  57. [57]

    Hao Wang, Daqing Zhang, Yasha Wang, Junyi Ma, Yuxiang Wang, and Shengjie Li. 2016. RT-Fall: A real-time and contactless fall detection system with commodity WiFi devices.IEEE Transactions on Mobile Computing16, 2 (2016), 511–526

    work page 2016

  58. [58]

    Xueyi Wang, Estefania Talavera, Dimka Karastoyanova, and George Azzopardi. 2023. Fall detection with a non-intrusive and first-person vision approach.IEEE Sensors Journal(2023)

    work page 2023

  59. [59]

    Yuxi Wang, Kaishun Wu, and Lionel M Ni. 2016. Wifall: Device-free fall detection by wireless networks.IEEE Transactions on Mobile Computing16, 2 (2016), 581–594

    work page 2016

  60. [60]

    2009.Biomechanics and motor control of human movement

    David A Winter. 2009.Biomechanics and motor control of human movement. John wiley & sons

    work page 2009

  61. [61]

    Sijie Yan, Yuanjun Xiong, and Dahua Lin. 2018. Spatial temporal graph convolutional networks for skeleton-based action recognition. In Proceedings of the AAAI conference on artificial intelligence, Vol. 32

    work page 2018

  62. [62]

    Lei Yu, Han Chen, Hantao He, Hong Nie, Xuping Zhai, and Bangshu Xiong. 2020. A fall detection system based on a thermopile imaging array and a back projection algorithm. In2020 IEEE International Conference on Electro Information Technology (EIT). IEEE, 060–065

    work page 2020

  63. [63]

    Xiaoqun Yu, Jaehyuk Jang, and Shuping Xiong. 2021. A large-scale open motion dataset (KFall) and benchmark algorithms for detecting pre-impact fall of the elderly using wearable inertial sensors.Frontiers in Aging Neuroscience13 (2021), 692865

    work page 2021

  64. [64]

    Sania Zahan, Ghulam Mubashar Hassan, and Ajmal Mian. 2022. Sdfa: Structure-aware discriminative feature aggregation for efficient human fall detection in video.IEEE Transactions on Industrial Informatics19, 8 (2022), 8713–8721

    work page 2022

  65. [65]

    Duo Zhang, Xusheng Zhang, Shengjie Li, Yaxiong Xie, Yang Li, Xuanzhi Wang, and Daqing Zhang. 2023. Lt-fall: The design and implementation of a life-threatening fall detection and alarming system.Proceedings of the ACM on Interactive, Mobile, Wearable and Ubiquitous Technologies7, 1 (2023), 1–24

    work page 2023

  66. [66]

    Jin Zhang, Cheng Wu, and Yiming Wang. 2020. Human fall detection based on body posture spatio-temporal evolution.Sensors20, 3 (2020), 946

    work page 2020

  67. [67]

    Xuehan Zhang, Zhongxu Bao, Yuqing Yin, Xu Yang, Xiao Xu, and Qiang Niu. 2023. LoFall: LoRa-based long-range through-wall fall detection. In2023 IEEE Symposium on Computers and Communications (ISCC). IEEE, 831–836. 24 TaFall Arxiv, 2026

    work page 2023

  68. [68]

    Xie Zhang, Chengxiao Li, and Chenshu Wu. 2025. TAPOR: 3D Hand Pose Reconstruction with Fully Passive Thermal Sensing for Around-Device Interactions.Proceedings of the ACM on Interactive, Mobile, Wearable and Ubiquitous Technologies9, 2 (2025), 1–25

    work page 2025

  69. [69]

    Xie Zhang and Chenshu Wu. 2024. TADAR: Thermal Array-based Detection and Ranging for Privacy-Preserving Human Sensing. In Proceedings of the Twenty-fifth International Symposium on Theory, Algorithmic Foundations, and Protocol Design for Mobile Networks and Mobile Computing. 11–20

    work page 2024

  70. [70]

    HongTao Zheng, Yan Liu, XiaoLi Wu, and YongHua Zhang. 2022. Realization of elderly fall integration monitoring system based on AlphaPose and YOLOV4. In2022 Asia Conference on Algorithms, Computing and Machine Learning (CACML). IEEE, 604–620

    work page 2022

  71. [71]

    Cankun Zhong, Wing WY Ng, Shuai Zhang, Chris D Nugent, Colin Shewell, and Javier Medina-Quero. 2020. Multi-occupancy fall detection using non-invasive thermal vision sensor.IEEE Sensors Journal21, 4 (2020), 5377–5388. 25

    work page 2020