Pith. sign in

REVIEW 5 major objections 6 minor 44 references

P2MFDS: A Privacy-Preserving Multimodal Fall Detection System for Elderly People in Bathroom Environments

T0 review · 5 major / 6 minor · reviewed 2026-08-06 · deepseek-v4-flash

Pith's one-line read Fusing macro-scale radar motion with micro-scale vibration impacts lets a privacy-preserving bathroom fall detector reach 95.0% accuracy.

desk verdict Useful dataset and a plausible sensor pairing, but the missing train/test split makes the headline accuracy unverifiable. read the letter →

arxiv 2506.17332 v1 pith:VRLN2Q7O submitted 2025-06-19 cs.CV cs.AI

classification cs.CVcs.AI
keywords falldetectionmmWaveradarvibrationsensingmultimodalfusionprivacy-preservingCNN-BiLSTM-Attentionbathroommonitoringelderlycare
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 tries to establish that combining millimeter-wave radar with a 3D vibration sensor, selected through a weighted scoring framework, can detect falls in a bathroom accurately while keeping the monitoring privacy-preserving. The proposed P2MFDS system feeds radar point clouds through a CNN–BiLSTM–Attention branch and vibration signals through a multi-scale CNN–SEBlock–Self-Attention branch, then fuses the two embeddings for a fall/non-fall decision. On a new dataset collected in a simulated bathroom with 48 adults performing scripted falls and daily activities, the paper reports 95.0% accuracy, 94.6% precision, 87.8% recall, and a 91.3% F1-score, outperforming 16 comparison systems. If reliable, this matters because bathrooms are where most elderly falls happen, and existing WiFi-, infrared-, or radar-only systems tend to degrade from environmental interference.

What carries the argument

The load-bearing mechanism is the dual-stream P2MFDS network plus the sensor-selection rule that picks its inputs. A weighted evaluation framework scores 14 sensing modalities on eight criteria (target relevance, non-intrusiveness, energy efficiency, computational complexity, deployability, recall, availability, cost-effectiveness) and selects mmWave radar plus triaxial vibration as the highest-scoring pair. The radar stream uses parallel 1D convolutions, a bidirectional LSTM, and attention to encode long-term motion; the vibration stream uses multi-scale convolutions, a Squeeze-and-Excitation channel-recalibration block, and self-attention to isolate short impact events. Their concatenated embeddings go through a lightweight detection head that outputs the fall/non-fall decision.

What would settle it

Take the released dataset and re-run the evaluation with leave-one-participant-out cross-validation, training on 47 people and testing on the remaining person; if accuracy drops well below the reported 95%, the original numbers were inflated by split leakage or by the simulated setting. A field test with elderly participants in real bathrooms would settle transferability directly.

Watch

Extended reading notes

Core claim

The central discovery is that a privacy-preserving fall detector does not need to choose between sensing the whole body and sensing the floor: it can do both. P2MFDS couples a radar branch that tracks macro-scale motion dynamics (velocity, distance, energy in 3D point clouds) with a vibration branch that captures micro-scale impact signatures, and concatenates the two learned embeddings before classification. On the paper's eight-scenario bathroom dataset, this fusion yields an overall accuracy of 95.0%, precision of 94.6%, recall of 87.8%, and F1-score of 91.3%, with per-scenario F1 above 90% in six of the eight activities and zero false negatives for squatting. The authors attribute the gain to complementarity: radar sees the fall as a movement pattern, vibration sees it as a mechanical impact, and each modality covers the other's blind spots under noise such as object drops, multipath reflections, and temperature shifts.

Load-bearing premise

The claim stands or falls on whether scripted falls by 48 healthy young adults in one simulated bathroom, with an unreported train/test split, are a faithful stand-in for real elderly falls, because leakage or demographic mismatch would inflate the reported 95%.

Editorial extensions

If this is right

  • If the reported numbers hold, bathrooms can be monitored for falls without cameras or wearables, preserving visual privacy while still triggering alerts.
  • The radar-plus-vibration pairing should suppress false alarms from non-fall events like soap or mop drops, since the two modalities must agree on both a motion pattern and an impact signature.
  • The public release of the dataset gives other researchers a common benchmark for privacy-preserving bathroom fall detection and a direct way to check the comparison numbers.
  • The architecture itself, long-term motion context from radar plus short-transient impact features from vibration with concatenation, provides a template for other multimodal health-monitoring tasks.

Reading between the lines

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

  • The eight-criteria weighted scoring framework is reusable beyond this paper: the same scores could rank sensing modalities for other private spaces or other health-monitoring tasks before any data collection.
  • Because recall (87.8%) trails precision (94.6%), a deployment worried more about missed falls than false alarms could lower the decision threshold, accepting extra false positives to catch more true falls, a tradeoff the paper does not examine.
  • The authors' own closing caveats, temperature fluctuations, partial occlusions, and the need for long-term deployments with elderly people, delimit the claim: the 95% figure is demonstrated only for a controlled, simulated bathroom, not yet for real elderly homes.
  • The macro/micro fusion pattern may transfer to other paired sensors, for example UWB radar with floor vibration or WiFi CSI with an accelerometer, wherever one modality sees the body and another sees the contact event.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

5 major / 6 minor

Summary. The paper proposes P2MFDS, a privacy-preserving fall detection system for bathrooms that fuses mmWave radar point clouds with triaxial vibration signals. The authors introduce a hand-scored sensor evaluation framework to justify the modality choice, collect a dataset from 48 healthy adults in a simulated bathroom, and design a dual-stream network whose embeddings are concatenated for fall/non-fall classification. They report 95.0% accuracy, 94.6% precision, 87.8% recall, and 91.3% F1-score, and compare against 16 prior methods. The paper promises public release of the dataset, code, and pretrained models, while also acknowledging in the Conclusion that large-scale long-term deployment with elderly people is still needed.

Significance. If the evaluation were validated, this system would be practically valuable: it uses low-cost, non-intrusive sensors, preserves privacy in a sensitive environment, and reports strong accuracy. The authors have taken care to collect data in a realistic bathroom, obtain informed consent and ethics approval, and state an intention to release data and code. However, the central performance claim is currently not trustworthy because the paper does not document a subject-independent evaluation protocol, the reported metrics are internally inconsistent, and the comparison with prior work is cross-dataset. The architecture description also contains a clear modality swap that prevents reproduction. These are load-bearing issues for the headline claim.

major comments (5)
  1. [Section IV.B and IV.C] The evaluation protocol never specifies a train/validation/test split or cross-validation scheme, so the reported 95.0% accuracy may arise from person-specific leakage. Since each of the 48 participants performs multiple trials across all scenarios, a random per-trial split could place the same person's data in both training and test sets. Given that radar point clouds and vibration signatures are person-specific, a subject-disjoint or leave-one-subject-out evaluation is needed to support the generalization claim to elderly users, who are absent from the training data.
  2. [Table III, Table IV, Table V, and Section IV.D] The reported metrics are internally inconsistent: Section IV.D gives precision 94.6%, recall 87.8%, and F1 91.3%, while Table III's total row for fall metrics lists recall 94.6%, precision 87.8%, and F1 91.3%, and Table V's full model lists recall 94.6%, precision 87.8%, and F1 91.1%; the Abstract and Introduction state recall 87.9%. As printed, the F1 of 91.3% is not consistent with P=0.946 and R=0.878, which give 91.1%. The authors must correct the metric definitions, regenerate the tables, and state unambiguously which number is precision and which is recall.
  3. [Section III.B, III.C, and Fig. 2] The architecture description is self-contradictory. The text at the start of Section III assigns the CNN-BiLSTM-Attention branch to radar and the Multi-Scale CNN-SEBlock-Self-Attention branch to vibration, but Section III.B describes the CNN-BiLSTM-Attention pipeline for vibration data, and Section III.C describes the multi-scale CNN-SEBlock-Self-Attention pipeline for radar data. This swap, together with the undefined F_acc in Eq. (8), makes the proposed network impossible to reproduce as written.
  4. [Table IV] The comparison with 16 prior methods is cross-dataset, using numbers taken from other papers evaluated on different sensors, environments, and protocols. The claim of significant gains over state-of-the-art is therefore not supported by the presented evidence. A same-protocol baseline or a controlled re-implementation is required to support the superiority claim, especially given the inconsistent metric labels in Table III and Table IV.
  5. [Section IV.B and Fig. 6] The dataset description is internally inconsistent: the text states nine experimental scenarios, while Fig. 6 and Table III refer to eight. The reported data volumes of 120,000 vibration points and 18,000 mmWave frames over 3 hours are far smaller than the sampling rates imply (approximately 1.08 million vibration samples and 108,000 radar frames at 100 Hz and 10 Hz, respectively). The authors should clarify the scenario count and reconcile the recorded data size with the stated sampling rates and trial durations.
minor comments (6)
  1. [Abstract and Section I] The abstract contains a sentence fragment: 'As aging is closely associated with increased fall risk, particularly in wet and confined environments such as bathrooms where over 80 percent of falls occur.' The recall value also differs between the Abstract (87.9%) and Section IV.D (87.8%).
  2. [Eq. (4)] The attention weight formula uses ht Wa without specifying whether Wa is a vector or a matrix, and the notation does not clearly define how a hidden-state vector produces a scalar score.
  3. [Eq. (8)] The fusion equation uses F_acc, which is not defined anywhere in the manuscript; it should presumably be F_vibration.
  4. [Section II, Table II] The weights w_i used in the overall usability score are not specified in the text or the table, so the sensor evaluation framework is not reproducible and its sensitivity to the chosen weights is not discussed.
  5. [Section III and Fig. 2] No hyperparameters are reported (kernel sizes, hidden units, learning rate, batch size, epochs, optimizer), so the network implementation cannot be independently reproduced.
  6. [Section IV.A and Fig. 5] The floor plan dimensions in Fig. 4 (2.50 m x 1.90 m) differ from the text's 2.5 m x 1.1 m x 2.2 m description; the floor area should be reconciled.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity found: the reported metrics are empirical outcomes on a collected dataset, not reductions to fitted inputs or self-cited premises.

full rationale

I find no significant circularity in this paper. The central claims are empirical measurements: Section IV.D reports an overall accuracy of 95.0%, precision of 94.6%, recall of 87.8%, and F1-score of 91.3% obtained by evaluating the trained P2MFDS network on the collected multimodal dataset. These numbers are not derived by construction from the sensor-evaluation framework or from any fitted parameter renamed as a prediction. The sensor evaluation framework in Section II (Table II) assigns hand-specified weights and scores to modalities and computes OUm = sum(w_i * S_{m,i}); selecting mmWave radar and 3D vibration because they receive the highest usability scores is an explicit design choice, and no subsequent reported metric is algebraically determined by those scores. The network equations (1)-(8) define standard filtering, feature extraction, attention, and fusion operations with learnable parameters; the reported test metrics are outcomes after training, not identities that hold by definition. No load-bearing argument reduces to a self-citation: the cited baselines in Table IV are external works, and the references do not supply a uniqueness theorem or ansatz that the paper depends on. The paper even states a limitation: 'large-scale, long-term deployments in real homes with elderly people are needed to further validate robustness and usability' (Section V), which is a validity caveat rather than a circular step. The main risks—absence of a documented train/test split, use of young adults simulating elderly falls, cross-dataset comparisons, and the F1 arithmetic inconsistency between 91.3% and 91.1%—are correctness, reproducibility, and reporting concerns, not circularity. I therefore assign score 0.

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

The system depends on several unstated values and assumptions: smoothing parameters, network hyperparameters, and the validity of simulated falls by young adults. No invented physical entities are introduced.

free parameters (4)
  • Sensor evaluation weights w_i = w = [3,2,3,2,2,3,2,1] (Table II)
    Used in OUm = sum(w_i * S_i) to select mmWave radar and vibration; values chosen by authors without sensitivity analysis.
  • Moving average window N (Eq. 1) = unspecified
    Controls vibration smoothing; no value reported.
  • Exponential low-pass filter alpha (Eq. 2) = unspecified
    Controls radar smoothing; no value reported.
  • Network hyperparameters (kernel sizes, hidden units, learning rate, batch size, epochs) = unspecified
    Architecture named but training details and hyperparameters omitted; these materially affect the reported results.
assumptions (4)
  • domain assumption Simulated falls by 20 to 30 year old adults are representative of elderly bathroom falls
    Section IV.C explicitly excludes elderly people for safety; conclusions nevertheless target elderly fall detection.
  • domain assumption mmWave radar point clouds and triaxial floor vibration carry sufficient discriminative information for fall detection in bathrooms
    Section III builds the whole network on this premise; no class-separability analysis is provided.
  • ad hoc to paper The hand-weighted sensor evaluation framework is a valid basis for selecting sensing modalities
    Table II weights and Table I scores are author-chosen; no external validation or robustness analysis.
  • domain assumption Data are partitioned into train and test sets without subject or scenario leakage
    Section IV describes data collection but never specifies the train/test split or cross-validation scheme.

how reviews work

0 comments
Cite this review

Pith. "Pith review of P2MFDS: A Privacy-Preserving Multimodal Fall Detection System for Elderly People in Bathroom Environments." pith.science (2026). https://pith.science/paper/VRLN2Q7O

@misc{pith2026250617332,
  author       = {Pith},
  title        = {Pith review of: P2MFDS: A Privacy-Preserving Multimodal Fall Detection System for Elderly People in Bathroom Environments},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/VRLN2Q7O}},
  note         = {Machine review of arXiv:2506.17332}
}
read the original abstract

By 2050, people aged 65 and over are projected to make up 16 percent of the global population. As aging is closely associated with increased fall risk, particularly in wet and confined environments such as bathrooms where over 80 percent of falls occur. Although recent research has increasingly focused on non-intrusive, privacy-preserving approaches that do not rely on wearable devices or video-based monitoring, these efforts have not fully overcome the limitations of existing unimodal systems (e.g., WiFi-, infrared-, or mmWave-based), which are prone to reduced accuracy in complex environments. These limitations stem from fundamental constraints in unimodal sensing, including system bias and environmental interference, such as multipath fading in WiFi-based systems and drastic temperature changes in infrared-based methods. To address these challenges, we propose a Privacy-Preserving Multimodal Fall Detection System for Elderly People in Bathroom Environments. First, we develop a sensor evaluation framework to select and fuse millimeter-wave radar with 3D vibration sensing, and use it to construct and preprocess a large-scale, privacy-preserving multimodal dataset in real bathroom settings, which will be released upon publication. Second, we introduce P2MFDS, a dual-stream network combining a CNN-BiLSTM-Attention branch for radar motion dynamics with a multi-scale CNN-SEBlock-Self-Attention branch for vibration impact detection. By uniting macro- and micro-scale features, P2MFDS delivers significant gains in accuracy and recall over state-of-the-art approaches. Code and pretrained models will be made available at: https://github.com/HaitianWang/P2MFDS-A-Privacy-Preserving-Multimodal-Fall-Detection-Network-for-Elderly-Individuals-in-Bathroom.

Figures

Figures reproduced from arXiv: 2506.17332 by the authors.

Figure 1
Figure 1. System overview of the P2MFDS pipeline. First, a multidimensional evaluation framework scores candidate sensing modalities (e.g., mmWave [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. Overview of the P2MFDS Network architecture. The upper pipeline employs a 1D CNN-BiLSTM-Attention sequence to extract global motion [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 3
Figure 3. Architecture of the SE Block. Channel-wise statistics are extracted [PITH_FULL_IMAGE:figures/full_fig_p004_3.png] view at source ↗
Figures from the paper (4 more)
Figure 4
Figure 4. Figure 4: Top-down floor plan of the experimental bathroom (2.50 m [PITH_FULL_IMAGE:figures/full_fig_p004_4.png]
Figure 5
Figure 5. Figure 5: Illustration of the controlled bathroom environment and sensor [PITH_FULL_IMAGE:figures/full_fig_p005_5.png]
Figure 6
Figure 6. Figure 6: Overview of the eight experimental scenarios simulating typical [PITH_FULL_IMAGE:figures/full_fig_p005_6.png]
Figure 7
Figure 7. Figure 7: Confusion matrices of the P2MFDS Network across eight ex [PITH_FULL_IMAGE:figures/full_fig_p006_7.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

44 extracted references · 42 canonical work pages

  1. [1]

    World population prospects 2019. department of economic and social affairs population dynamics,

    D. Leong, K. Teo, S. Rangarajan, P. Lopez-Jaramillo, A. Avezum Jr, A. Orlandini et al. , “World population prospects 2019. department of economic and social affairs population dynamics,” World, vol. 73, no. 7, 2018

  2. [2]

    Fall risk assessment scales: A systematic literature review,

    V . Strini, R. Schiavolin, and A. Prendin, “Fall risk assessment scales: A systematic literature review,” Nursing Reports , vol. 11, no. 2, pp. 430–443, 2021

  3. [3]

    Elderly fall detection systems: A literature survey,

    X. Wang, J. Ellul, and G. Azzopardi, “Elderly fall detection systems: A literature survey,” Frontiers in Robotics and AI , vol. 7, p. 71, 2020

  4. [4]

    Sensor technologies for fall detection systems: A review,

    A. Singh, S. U. Rehman, S. Yongchareon, and P. H. J. Chong, “Sensor technologies for fall detection systems: A review,” IEEE Sensors Journal, vol. 20, no. 13, pp. 6889–6919, 2020

  5. [5]

    Multi-sensor ambient assisted liv- ing system for fall detection,

    A. Yazar, F. Erden, and A. E. Cetin, “Multi-sensor ambient assisted liv- ing system for fall detection,” in Proceedings of the IEEE international conference on acoustics, speech, and signal processing (ICASSP’14) , 2014, pp. 1–3

  6. [6]

    Elderly patient monitoring and fall detection using mmwave fmcw radar system,

    M. S. R. Akash, M. N. Shahria, M. A. Morshed, S. S. Rodsee, N. Hannan, and M. H. Imam, “Elderly patient monitoring and fall detection using mmwave fmcw radar system,” in 2023 26th Interna- tional Conference on Computer and Information Technology (ICCIT) . IEEE, 2023, pp. 1–6

  7. [7]

    Fall detection based on key points of human-skeleton using openpose,

    W. Chen, Z. Jiang, H. Guo, and X. Ni, “Fall detection based on key points of human-skeleton using openpose,” Symmetry, vol. 12, no. 5, p. 744, 2020

  8. [8]

    Internet of things (iot) applications for elderly care: a reflective review,

    S. Y . Y . Tun, S. Madanian, and F. Mirza, “Internet of things (iot) applications for elderly care: a reflective review,” Aging clinical and experimental research, vol. 33, pp. 855–867, 2021

Show all 44 references
  1. [9]

    Care needs of the elderly who live alone: an intersectoral perception,

    G. P. Cardoso, D. G. Damaceno, M. F. S. Alarcon, and M. J. S. Marin, “Care needs of the elderly who live alone: an intersectoral perception,” Rev. Rene, vol. 21, pp. 44 395–44 395, 2020

  2. [10]

    Moving average and savitzki-golay smoothing filters using mathcad,

    J. L. Gui ˜n´on, E. Ortega, J. Garc ´ıa-Ant´on, and V . P ´erez-Herranz, “Moving average and savitzki-golay smoothing filters using mathcad,” Papers ICEE, vol. 2007, pp. 1–4, 2007

  3. [11]

    A review on fall detection systems in bathrooms: challenges and opportunities,

    T. Perumal et al., “A review on fall detection systems in bathrooms: challenges and opportunities,” Multimedia Tools and Applications, pp. 1–29, 2024

  4. [12]

    Falldefi: Ubiquitous fall detection using commodity wi-fi devices,

    S. Palipana, D. Rojas, P. Agrawal, and D. Pesch, “Falldefi: Ubiquitous fall detection using commodity wi-fi devices,”Proceedings of the ACM on Interactive, Mobile, Wearable and Ubiquitous Technologies, vol. 1, no. 4, pp. 1–25, 2018

  5. [13]

    ehomeseniors dataset: An infrared thermal sensor dataset for automatic fall detection research,

    F. Riquelme, C. Espinoza, T. Rodenas, J.-G. Minonzio, and C. Tara- masco, “ehomeseniors dataset: An infrared thermal sensor dataset for automatic fall detection research,” Sensors, vol. 19, no. 20, p. 4565, 2019

  6. [14]

    Wagner, P

    J. Wagner, P. Mazurek, and R. Z. Morawski, Non-invasive monitoring of elderly persons: Systems based on impulse-radar sensors and depth sensors. Springer, 2022

  7. [15]

    Millimeter-wave energy harvesting using 4 x 4 microstrip patch antenna array,

    A. Mavaddat, S. H. M. Armaki, and A. R. Erfanian, “Millimeter-wave energy harvesting using 4 x 4 microstrip patch antenna array,” IEEE Antennas and wireless propagation letters, vol. 14, pp. 515–518, 2014

  8. [16]

    Highly efficient environment for fmcw radar phase noise simulations in if domain,

    M. Gerstmair, A. Melzer, A. Onic, R. Stuhlberger, and M. Huemer, “Highly efficient environment for fmcw radar phase noise simulations in if domain,” IEEE Transactions on Circuits and Systems II: Express Briefs, vol. 65, no. 5, pp. 582–586, 2018

  9. [17]

    A method for automatic fall detection of elderly people using floor vibrations and sound—proof of concept on human mimicking doll falls,

    Y . Zigel, D. Litvak, and I. Gannot, “A method for automatic fall detection of elderly people using floor vibrations and sound—proof of concept on human mimicking doll falls,” IEEE transactions on biomedical engineering, vol. 56, no. 12, pp. 2858–2867, 2009

  10. [18]

    An analysis of the classical doppler effect,

    C. Neipp, A. Hernndez, J. Rodes, A. Mrquez, T. Belndez, and A. Belndez, “An analysis of the classical doppler effect,” European journal of physics , vol. 24, no. 5, p. 497, 2003

  11. [19]

    Intelligent ensemble learning approach for phishing website detection based on weighted soft voting,

    A. Taha, “Intelligent ensemble learning approach for phishing website detection based on weighted soft voting,” Mathematics, vol. 9, no. 21, p. 2799, 2021

  12. [20]

    Fall detection system based on real-time pose estimation and svm,

    Y . Chen, R. Du, K. Luo, and Y . Xiao, “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, 2021, pp. 990–993

  13. [21]

    Accurate fall detection for patients with parkinson’s disease based on a data event algorithm and wireless sensor nodes,

    H. A. Hashim, S. L. Mohammed, and S. K. Gharghan, “Accurate fall detection for patients with parkinson’s disease based on a data event algorithm and wireless sensor nodes,” Measurement, vol. 156, p. 107573, 2020

  14. [22]

    Falls and fall prevention in the elderly,

    M. K. Appeadu and B. Bordoni, “Falls and fall prevention in the elderly,” in StatPearls [Internet]. StatPearls Publishing, 2023

  15. [23]

    Classification of emg signal based on time domain and frequency domain features,

    J. Too, A. R. Abdullah, T. T. Zawawi, N. M. Saad, and H. Musa, “Classification of emg signal based on time domain and frequency domain features,” International Journal of Human and Technology Interaction (IJHaTI), vol. 1, no. 1, pp. 25–30, 2017

  16. [24]

    Wireless power transfer techniques for implantable medical devices: A review,

    S. R. Khan, S. K. Pavuluri, G. Cummins, and M. P. Desmulliez, “Wireless power transfer techniques for implantable medical devices: A review,” Sensors, vol. 20, no. 12, p. 3487, 2020

  17. [25]

    A systematic review of the guide- lines and delphi study for the multifactorial fall risk assessment of community-dwelling elderly,

    J. Kim, W. Lee, and S. H. Lee, “A systematic review of the guide- lines and delphi study for the multifactorial fall risk assessment of community-dwelling elderly,” International journal of environmental research and public health , vol. 17, no. 17, p. 6097, 2020

  18. [26]

    Squeeze-and-excitation networks,

    J. Hu, L. Shen, and G. Sun, “Squeeze-and-excitation networks,” in Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, 2018, pp. 7132–7141

  19. [27]

    1d convolutional neural networks and applications: A survey,

    S. Kiranyaz, O. Avci, O. Abdeljaber, T. Ince, M. Gabbouj, and D. J. Inman, “1d convolutional neural networks and applications: A survey,” Mechanical systems and signal processing , vol. 151, p. 107398, 2021

  20. [28]

    Review of deep learning: concepts, cnn architectures, challenges, applications, future directions,

    L. Alzubaidi, J. Zhang, A. J. Humaidi, A. Al-Dujaili, Y . Duan, O. Al- Shamma, J. Santamar´ıa, M. A. Fadhel, M. Al-Amidie, and L. Farhan, “Review of deep learning: concepts, cnn architectures, challenges, applications, future directions,” Journal of big Data , vol. 8, pp. 1– 74, 2021

  21. [29]

    Unobtrusive human fall detection system using mmwave radar and data driven methods,

    A. Rezaei, A. Mascheroni, M. C. Stevens, R. Argha, M. Papandrea, A. Puiatti, and N. H. Lovell, “Unobtrusive human fall detection system using mmwave radar and data driven methods,” IEEE Sensors Journal, vol. 23, no. 7, pp. 7968–7976, 2023

  22. [30]

    Fall detection with uwb radars and cnn-lstm architecture,

    J. Maitre, K. Bouchard, and S. Gaboury, “Fall detection with uwb radars and cnn-lstm architecture,” IEEE Journal of Biomedical and Health Informatics, vol. 25, no. 4, pp. 1273–1283, 2021

  23. [31]

    Fall detection system using millimeter-wave radar based on neural network and information fusion,

    Y . Yao, C. Liu, H. Zhang, B. Yan, P. Jian, P. Wang, L. Du, X. Chen, B. Han, and Z. Fang, “Fall detection system using millimeter-wave radar based on neural network and information fusion,” IEEE Internet of Things Journal , vol. 9, no. 21, pp. 21 038–21 050, 2022

  24. [32]

    Real-time fall detection using mmwave radar,

    W. Li, D. Zhang, Y . Li, Z. Wu, J. Chen, D. Zhang, Y . Hu, Q. Sun, and Y . Chen, “Real-time fall detection using mmwave radar,” in ICASSP 2022 - 2022 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP) , 2022, pp. 16–20

  25. [33]

    A noncontact fall detection method for bedside application with a mems infrared sensor and a radar sensor,

    C. He, S. Liu, G. Zhong, H. Wu, L. Cheng, G. Yan, and Y . Wen, “A noncontact fall detection method for bedside application with a mems infrared sensor and a radar sensor,” IEEE Internet of Things Journal , vol. 10, no. 14, pp. 12 577–12 589, 2023

  26. [34]

    Smart seismic sensing for indoor fall detection, location, and notification,

    J. Clemente, F. Li, M. Valero, and W. Song, “Smart seismic sensing for indoor fall detection, location, and notification,” IEEE Journal of Biomedical and Health Informatics, vol. 24, no. 2, pp. 524–532, 2020

  27. [35]

    Elderly fall detection with vital signs monitoring using cw doppler radar,

    K. Hanifi and M. E. Karsligil, “Elderly fall detection with vital signs monitoring using cw doppler radar,” IEEE Sensors Journal , vol. 21, no. 15, pp. 16 969–16 978, 2021

  28. [36]

    Non-contact fall detection system based on mmwave radar and transformer,

    J. Wang, L.-H. Lin, M. Wang, and W. Xu, “Non-contact fall detection system based on mmwave radar and transformer,” IEEE Sensors Journal, vol. 23, no. 20, pp. 24 002–24 009, 2023

  29. [37]

    Compressed domain contact- less fall incident detection using uwb radar signals,

    H. Sadreazami, M. Bolic, and S. Rajan, “Compressed domain contact- less fall incident detection using uwb radar signals,” IEEE Transac- tions on Biomedical Engineering , vol. 69, no. 1, pp. 244–256, 2022

  30. [38]

    Radar-based elderly fall detection using spwvd and resnet network,

    P. Swarubini and N. Ganapathy, “Radar-based elderly fall detection using spwvd and resnet network,” Current Directions in Biomedical Engineering, vol. 10, no. 4, pp. 498–501, 2024

  31. [39]

    An intelligent fall detection method based on millimeter-wave radar and hybrid deep learning model,

    B. Yang, L. Guo, X. Zhang, J. Yang, Y . Ren, and X. Li, “An intelligent fall detection method based on millimeter-wave radar and hybrid deep learning model,” Sensors, vol. 22, no. 21, p. 8176, 2022

  32. [40]

    Multimodal fusion for indoor fall detection using millimeter-wave radar and vision sensor,

    J. Dai, Z. Ji, Z. Xie, C. Wang, and J. Wu, “Multimodal fusion for indoor fall detection using millimeter-wave radar and vision sensor,” Expert Systems with Applications , vol. 231, p. 120770, 2023

  33. [41]

    Attention-gru based multimodal fusion for elderly fall detection,

    Y . Sun, G. Li, Z. Zhu, C. Huang, X. Hou, and Y . He, “Attention-gru based multimodal fusion for elderly fall detection,” Sensors, vol. 21, no. 2, pp. 1–16, 2021

  34. [42]

    Multimodal sensor fusion for elderly fall detection using wi-fi csi and accelerometer data,

    X. Meng, Y . Liu, B. Zhou, P. Wang, J. Song, and D. Zhang, “Multimodal sensor fusion for elderly fall detection using wi-fi csi and accelerometer data,” IEEE Internet of Things Journal , vol. 9, no. 21, pp. 21 247–21 257, 2022

  35. [43]

    Fall detection using radar and deep learning for elderly safety,

    H. Alkhaldi and M. A. A. Faruque, “Fall detection using radar and deep learning for elderly safety,” in Proc. IEEE Sensors Applications Symposium (SAS). IEEE, 2022, pp. 1–6

  36. [44]

    A privacy-preserving radar- based system for elderly fall detection using attention-enhanced cnn- lstm architecture,

    Y . Zhang, L. Wang, M. Li, and X. Liu, “A privacy-preserving radar- based system for elderly fall detection using attention-enhanced cnn- lstm architecture,” in Proceedings of the 45th Annual International Conference of the IEEE Engineering in Medicine and Biology Society (EMB...

Pith tools

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