Pith. sign in

REVIEW 4 major objections 6 minor 53 references

Securing Contrastive mmWave-based Human Activity Recognition against Adversarial Label Flipping

T0 review · 4 major / 6 minor · reviewed 2026-08-06 · deepseek-v4-flash

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

Pith's one-line read Trajectory-aware label flipping can severely degrade contrastive mmWave-based HAR, and a confidence-based defense keeps accuracy above 90 percent even at 40 percent poisoned labels.

desk verdict A useful, well-scoped empirical study of label flipping on contrastive mmWave HAR, with a promising defense whose reported robustness is under-evidenced because the confident-example filter is never validated. read the letter →

arxiv 2608.04029 v1 pith:2OJKUXCI submitted 2026-08-01 cs.CR

Amit Singha , Ziqian Bi , Tao Li , Yimin Chen , Yanchao Zhang This is my paper
classification cs.CR
keywords humanactivityrecognitionlabelflippingpoisoningmmWaveradarsensingsupervisedcontrastivelearningwirelesssecurityadversarialnoiseconfidentsampleselection
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

This paper tries to establish that label-flipping poisoning is a real threat to mmWave-based human activity recognition (HAR) when the recognizer is trained with supervised contrastive learning (SCL). It defines three trajectory-aware attacks — random flipping, flipping across dissimilar motion trajectories, and flipping within similar trajectories — and shows on a six-activity radar prototype that they all lower accuracy, with random flipping the most damaging and SCL more vulnerable than plain supervised learning. The paper then proposes Sel-CL, a defense that selects confident examples and confident pairs from the poisoned training set using unsupervised contrastive learning, without needing any trusted clean dataset. On its prototype, Sel-CL keeps accuracy nearly unchanged up to 20% poisoned labels, above 90% for all three attacks at 40% poisoning, and above 90% even at 80% poisoning under random flips, while also improving clean-data accuracy. If correct, this matters because wireless HAR is moving toward contrastive training, and its heatmap-like inputs make flipped labels much harder for a human to spot than flipped image labels.

What carries the argument

The load-bearing mechanism is Sel-CL's confident-example and confident-pair selection loop. After a few epochs of unsupervised contrastive learning, every training sample is embedded as a low-dimensional vector $\boldsymbol{z}_i$; each sample receives a pseudo-label from the majority class among its top-$K$ nearest neighbors, its per-class clean probability is approximated by the fraction of neighbors in each class, and an example is called confident when its cross-entropy against the poisoned label falls below a class-specific threshold. Confident pairs are pairs with matching confident labels and representation similarity above a threshold from a $\beta$-fractile, and only those pairs feed the supervised contrastive loss, while non-confident samples are handled with unsupervised contrastive loss; a mixup term blends pairs, and classification plus similarity losses on confident examples stabilize training, followed by fine-tuning with a robust cross-entropy loss. This selection loop carries the argument because every defense accuracy claim rests on it rejecting enough poisoned labels to keep the contrastive embedding space clean.

What would settle it

Poison only one trajectory-similar pair (for example, flip Push labels to Pull) at 40% while sweeping $K$ from 5 to 500 and the $\alpha$/$\beta$ thresholds; if Sel-CL accuracy falls to the unpoisoned SCL baseline or below in any cell, the claim that it stays above 90% for all attacks at 40% poison is falsified.

Watch

Extended reading notes

Core claim

The paper's central discovery is that label flipping is a practical attack surface for contrastive mmWave-based HAR and that attack power depends on trajectory similarity. On a TI 1843 radar dataset of 10,650 samples covering six hand activities collected from 25 volunteers, supervised contrastive learning reaches 97.92% test accuracy on clean data but degrades faster than the 96.43%-accurate supervised baseline when labels are flipped; random flips are the most effective, followed by inner-trajectory flips (similar motions) and cross-trajectory flips. The countermeasure, Sel-CL, first learns embeddings with unsupervised contrastive learning, assigns each sample a pseudo-label $\hat{y}_i$ by majority vote over its top-$K$ nearest neighbors ($K=250$), and keeps only confident examples and confident pairs, with per-class thresholds set by $\alpha$- and $\beta$-fractiles, for supervised contrastive training with mixup and a similarity loss. The paper reports that this keeps test accuracy above 90% for all three attacks at 40% poisoned labels, above 90% for random flips even at 80% poisoning, and that clean-data accuracy rises to 98.78%.

Load-bearing premise

The defense assumes that unsupervised contrastive learning on the poisoned data still puts true same-class examples among each other's top-$K$ nearest neighbors, so the majority pseudo-label recovers the clean label; if label flipping distorts those neighborhoods enough, confident selection fails and the accuracy guarantees collapse.

Editorial extensions

If this is right

  • Contrastive mmWave HAR models, despite their clean-data edge, are more vulnerable than plain supervised models to label-flipping poisoning, so deploying them without a defense raises security risk.
  • Random label flips are the strongest of the three studied attacks; fixed flips within or across trajectories are weaker but still drag accuracy down as the poison ratio grows, and symmetric flips are generally more damaging than asymmetric ones because they involve more classes.
  • If Sel-CL works as reported, a HAR operator can tolerate substantial label noise — up to 40% under all studied attacks and up to 80% under random flips — without needing a trusted clean dataset for calibration.
  • Sel-CL also raises clean-data accuracy to 98.78% from 97.92% for SCL and 96.43% for SL, so the defense does not sacrifice normal-case performance.
  • The attack and defense principles are stated to extend to other wireless HAR modalities, including Wi-Fi and acoustic sensing.

Reading between the lines

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

  • An adaptive adversary who knows the top-$K$ majority rule could concentrate flips inside one trajectory-similar pair, such as Push/Pull, to make the poisoned label the local neighborhood majority; the paper varies the poison ratio but not the flip distribution, so that regime is untested.
  • The defense's hyperparameters ($K$, $\alpha$, $\beta$) are fixed from stability observations; sweeping them under each attack would show whether the reported margins are a plateau or a knife-edge.
  • Because the selection mechanism works on learned embeddings rather than raw radar physics, Sel-CL may transfer to Wi-Fi CSI and acoustic HAR, but that transfer is an extrapolation beyond the experiments reported here.
  • Sel-CL's clean-data gain suggests confident-example selection could serve as a general label-noise filter for wireless sensing datasets, independent of adversarial intent.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

4 major / 6 minor

Summary. This paper studies label flipping poisoning attacks on millimeter-wave (mmWave) human activity recognition (HAR) systems trained with supervised contrastive learning (SCL). It introduces three attacks—random label flipping, across-trajectory flipping (between activities with dissimilar trajectories), and inner-trajectory flipping (between activities with similar trajectories, e.g., Push/Pull, Slide Left/Right, Clockwise/Anticlockwise)—and evaluates them on a self-built mmWave radar dataset with six hand activities. The paper further proposes a defense, Sel-CL, which uses unsupervised contrastive learning (UCL) to identify 'confident' examples and pairs from the poisoned dataset, then trains SCL on those selected examples. Experiments report that the attacks significantly degrade accuracy of both supervised and SCL baselines, and that Sel-CL maintains above 90% test accuracy up to 40% poisoned labels for all attacks, with clean-data accuracy of 98.78%. The paper claims to be the first systematic study of label flipping in mmWave-based HAR and emphasizes that the defense does not require a trusted dataset.

Significance. The paper addresses a real and previously underexplored security concern: poisoning of the training data used by contrastive-learning-based wireless HAR. The trajectory-aware attack taxonomy (random, across-, inner-trajectory) is a useful framing that captures the intuition that similarly moving activities are harder to distinguish, and the experimental prototype uses data from 25 volunteers across multiple environments with a domain-shift test set. The proposed defense is notable for avoiding the trusted-dataset assumption of prior label-noise defenses, instead relying on UCL embeddings and KNN pseudo-labels. If the robustness claims are empirically solid, this would be a meaningful contribution to the WiSec community. However, the current empirical validation is insufficient: only single-run accuracy curves are reported, no comparison to prior robust-training baselines is given, and the core confident-example selection mechanism is not validated with diagnostic metrics. The paper's contribution is thus conditional on additional experiments that, in principle, are within the manuscript's scope.

major comments (4)
  1. [Section 6.3, Figure 16] The central robustness claim—that Sel-CL maintains above 90% accuracy at 40% poisoned labels for all attacks—is supported only by single-run accuracy curves with no error bars, confidence intervals, or repeated independent runs. As a result, the reported margins (e.g., the differences among the attack curves in Figure 16 and the ordering of attack potency in Figure 14) cannot be distinguished from training noise. Please report means and standard deviations over at least five independent runs, and where applicable use paired significance tests across attack conditions.
  2. [Section 5.2, Eq. (4)] The confident-example filter defined by Eq. (4) is the mechanism that is claimed to remove poisoned labels, but the paper never reports diagnostics: the accuracy of the KNN pseudo-labels q̂(x_i), the precision/recall of the confident set T, or the fraction of flipped labels that survive selection. For the inner-trajectory attack (which flips between similar classes such as Push/Pull), it is not demonstrated that UCL embeddings place true same-class instances among the top-250 nearest neighbors; if they do not, Eq. (4) can exclude clean examples or admit poisoned ones, and the reported >90% accuracy under inner-trajectory attacks is not established. Please add these diagnostics, broken down by attack type and poisoning ratio.
  3. [Section 5.2 and Section 6] The defense has multiple hyperparameters: K (nearest neighbors for pseudo-labels), the α-percentile for class thresholds γ_c, the β-fractile for pair-similarity threshold γ, the Mixup Beta distribution parameters, the temperature θ in the contrastive losses, and the loss weights λ_c and λ_s. All experiments use a single default setting, and no sensitivity analysis is provided. The only justification is a sentence in Section 5.2 that the method's efficacy remains stable once K is sufficiently high, with no supporting data. Because the thresholds are derived from the poisoned training data itself, it is essential to show that the defense does not rely on finely tuned values that happen to work for this specific dataset. Please provide ablations over K, α, β, and θ.
  4. [Section 6.3] The evaluation does not compare Sel-CL to any prior defense or robust training baseline, even though the defense is explicitly adapted from selective-supervised contrastive learning (Ref. [15]) and multi-objective interpolation training (Ref. [23]). Without comparison to these baselines under the same attack settings, the reader cannot tell whether the observed robustness comes from the proposed modifications or from pre-existing methods. Please include at least the two cited baselines (Li et al. and Ortego et al.) evaluated under the same conditions.
minor comments (6)
  1. [Section 5.2] The definition of the α-percentile is unclear: the text says 'calculated as: Σ_i I[ŷ_i = ỹ_i] I[ỹ_i = c], c∈[6]' which is a per-class agreement count, not a percentile. Please clarify the exact procedure and report the α and β values used.
  2. [Section 4.2] The symmetric and asymmetric attack definitions are only described in prose. A formal definition with equations (e.g., which classes are flipped in which direction and how many samples) would improve precision and reproducibility.
  3. [Section 6.1] The data split (6,300 training, 2,100 validation, 2,250 test) sums to the stated 10,650, but the split by environment is described only in prose. A table listing the number of samples per environment and per class would be helpful.
  4. [Figures 6-9] Some figure captions have garbled or missing text in the extracted manuscript, and the confusion matrices lack a legend or color scale description, making them difficult to interpret. Please ensure the final figures are complete and self-explanatory.
  5. [Section 1 and Section 7] The paper claims 'first systematic study' twice; to avoid overclaiming, it would be useful to more explicitly delineate the novelty of the defense relative to Refs. [15] and [23], since the defense is an adaptation of these methods.
  6. [Section 5.2 and 5.3] The temperature θ in Eqs. (1) and (5) and the Mixup Beta distribution parameters are not reported. Please provide the exact values or a reference to the implementation.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity found: the paper's attack and defense claims are empirical, evaluated on a held-out environment test set, and no result reduces by construction to a fitted constant or self-citation.

full rationale

This is an empirical systems and security paper rather than a derivation, and the claimed chain (UCL representation, KNN pseudo-labels, confident-example filter, SCL training, fine-tuning, held-out test accuracy) contains no step in which a predicted quantity is defined as the output of a fit. The defense's pseudo-label aggregation in Eq. (3) and confidence filter in Eq. (4) use the given, possibly poisoned labels as inputs, but the reported accuracy is measured on the living-room test set, which was not used for training (bedroom, laboratory, and two office rooms) or validation (meeting room). The >90% accuracy claim is therefore an external empirical outcome, not a tautology. The paper explicitly adapts prior work [15, 23] for the defense and cites [4, 12, 37] for standard components; the only overlapping-author citations ([13, 41]) are not load-bearing for the attack or defense claims. The lack of sensitivity analysis for K=250 and the alpha/beta thresholds is a validation gap, not circularity, because those thresholds are not renamed as predictions and the central accuracy claims do not reduce to them by equation. Accordingly, no specific circular step can be quoted, and the score is 0.

Assumptions & free parameters 6 free parameters · 3 assumptions · 0 invented entities

The ledger shows that the central experimental claims rest on several underspecified hyperparameters and on two domain assumptions about embedding geometry under poisoning. No new physical or mathematical entities are introduced.

free parameters (6)
  • K (nearest neighbors for pseudo-labels) = 250
    Chosen in Section 5.2 with the claim that efficacy remains stable for sufficiently high K; no sensitivity analysis or plot is provided.
  • alpha-percentile for per-class threshold gamma_c = not specified
    Section 5.2 defines T_c via gamma_c set by the alpha-percentile of label agreements; the value of alpha is not reported.
  • beta-fractile for pair similarity threshold gamma = not specified
    Section 5.2 says gamma is determined by the beta-fractile of representation similarity scores; beta is not reported.
  • Mixup Beta distribution parameters = not specified
    Section 5.3 states lambda follows a Beta distribution but gives no parameters.
  • temperature theta in contrastive losses = not specified
    Equations (1) and (5) use theta as a temperature parameter; its value is never given.
  • loss weights lambda_c and lambda_s = lambda_c=1, lambda_s=0.01
    Set in Section 5.3 and used in all experiments; these are fixed, not tuned per attack.
assumptions (3)
  • domain assumption Trajectory similarity groups (Push/Pull, Slide Left/Right, Clockwise/Anticlockwise) correspond to representation similarity in the learned embedding space.
    Section 4.2 defines the three attacks on this grouping; if the grouping does not match embedding geometry, the inner-trajectory/cross-trajectory distinction loses meaning.
  • domain assumption Unsupervised contrastive learning on poisoned labels still produces embeddings where true same-class instances are nearest neighbors.
    Section 5.2 relies on this to compute pseudo-labels from top-K neighbors; no theoretical or empirical failure analysis is given.
  • domain assumption No trusted training dataset is needed because confident examples are identified from the poisoned data itself.
    This is the paper's stated advantage over [26,27]; it assumes the clean majority dominates the K-neighborhoods even at high poisoning ratios.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Securing Contrastive mmWave-based Human Activity Recognition against Adversarial Label Flipping." pith.science (2026). https://pith.science/paper/2OJKUXCI

@misc{pith2026260804029,
  author       = {Pith},
  title        = {Pith review of: Securing Contrastive mmWave-based Human Activity Recognition against Adversarial Label Flipping},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/2OJKUXCI}},
  note         = {Machine review of arXiv:2608.04029}
}
read the original abstract

Wireless Human Activity Recognition (HAR), leveraging their non-intrusive nature, has the potential to revolutionize various sectors, including healthcare, virtual reality, and surveillance. The advent of millimeter wave (mmWave) technology has significantly enhanced the capabilities of wireless HAR systems. This paper presents the first systematic study on the vulnerabilities of mmWave-based HAR to label flipping poisoning attacks in the context of supervised contrastive learning. We identify three label poisoning attacks on the contrastive mmWave-based HAR and propose corresponding countermeasures. The efficacy of the attacks and also our countermeasures are experimentally validated on a prototype system. The attacks and countermeasures can be easily extended to other wireless HAR systems, thereby promoting security considerations in system design and deployment.

Figures

Figures reproduced from arXiv: 2608.04029 by the authors.

Figure 1
Figure 1. Heatmaps of two activities. One breakthrough in this field is millimeter wave (mmWave) technology, which operates within the frequency range of 24 GHz to 300 GHz, corresponding to wavelengths between 12.5 mm and 1 mm. The exceptional bandwidth and high-speed capabilities of mmWave have unlocked new possibilities for HAR applications requiring low latency and high speeds over short distances [22, 46]. In addition to … view at source ↗
Figure 2
Figure 2. A basic mmWave-based HAR system. 2 A Basic mmWave-based HAR System In this section, we introduce the basic mmWave-based HAR system that employs supervised learning (SL) with a TI 1843 mmWave radar following the design in [16, 17]. During activity recognition, the transmission antennas of the radar first emit frequency-modulated continuous wave (FMCW) chirps. The signals are then reflected by each part of the user bo… view at source ↗
Figure 3
Figure 3. The CNN-LSTM network used for activity recognition. The system finally uses a hybrid CNN-LSTM model for activity classification, shown in [PITH_FULL_IMAGE:figures/full_fig_p002_3.png] view at source ↗
Figures from the paper (14 more)
Figure 4
Figure 4. Figure 4: Architecture of supervised contrastive learning [PITH_FULL_IMAGE:figures/full_fig_p003_4.png]
Figure 5
Figure 5. Figure 5: Architecture of the proposed defense against [PITH_FULL_IMAGE:figures/full_fig_p004_5.png]
Figure 6
Figure 6. Figure 6: Test confusion matrix of SL-based HAR. CT AT PL PH LS RS Predicted label CT AT PL PH LS RS True label 99.2 1.3 0.0 0.0 0.0 0.3 0.5 97.3 0.5 0.0 0.0 0.0 0.3 1.1 93.6 0.0 0.0 0.0 0.0 0.0 0.0 99.5 0.3 1.1 0.0 0.3 5.9 0.0 99.2 0.0 0.0 0.0 0.0 0.5 0.5 98.7 0 20 40 60 80 [P…
Figure 7
Figure 7. Figure 7: Test confusion matrix of SCL-based HAR. For our study, we divided the dataset into training, validation, and test sets based on the environment of data collection. The train￾ing set, comprising 6,300 samples, was derived from the bedroom, laboratory, and two office roo…
Figure 8
Figure 8. Figure 8: Test confusion matrix of SL-based HAR on [PITH_FULL_IMAGE:figures/full_fig_p007_8.png]
Figure 9
Figure 9. Figure 9: Test confusion matrix of SCL-based HAR on [PITH_FULL_IMAGE:figures/full_fig_p007_9.png]
Figure 10
Figure 10. Figure 10: Feature representations for SL-based HAR with clean training data and under random attacks. [PITH_FULL_IMAGE:figures/full_fig_p008_10.png]
Figure 11
Figure 11. Figure 11: Feature representations for SCL-based HAR with clean training data and under random attacks. [PITH_FULL_IMAGE:figures/full_fig_p008_11.png]
Figure 12
Figure 12. Figure 12: Feature representations for Sel-CL-based HAR with clean training data and under random attacks. [PITH_FULL_IMAGE:figures/full_fig_p008_12.png]
Figure 13
Figure 13. Figure 13: Impact of three label poisoning attacks on SL and SCL based HAR systems. [PITH_FULL_IMAGE:figures/full_fig_p009_13.png]
Figure 14
Figure 14. Figure 14: Comparing three attacks on SCL-based HAR. [PITH_FULL_IMAGE:figures/full_fig_p009_14.png]
Figure 15
Figure 15. Figure 15: Feature representations of SCL-based HAR under three different attacks with poisoned label ratio of 40%. [PITH_FULL_IMAGE:figures/full_fig_p010_15.png]
Figure 16
Figure 16. Figure 16: Performance of Sel-CL in defending against [PITH_FULL_IMAGE:figures/full_fig_p010_16.png]
Figure 17
Figure 17. Figure 17: Accuracy of three techniques under random [PITH_FULL_IMAGE:figures/full_fig_p010_17.png]

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

53 extracted references · 42 canonical work pages

  1. [15]

    Shikun Li, Xiaobo Xia, Shiming Ge, and Tongliang Liu. 2022. Selective-Supervised Contrastive Learning with Noisy Labels. InIEEE CVPR. New Orleans, LA

    work page 2022

  2. [23]

    O’Connor, and Kevin McGuinness

    Diego Ortego, Eric Arazo, Paul Albert, Noel E. O’Connor, and Kevin McGuinness

    work page

  3. [1]

    Görkem Algan and İlkay Ulusoy. 2020. Label Noise Types and Their Effects on Deep Learning. arXiv:2003.10471 [cs.CV]

    work page Pith review arXiv 2020

  4. [2]

    Anum Ali, Priyabrata Parida, Vutha Va, Saifeng Ni, Khuong Nhat Nguyen, Boon Loong Ng, and Jianzhong Charlie Zhang. 2022. End-to-End Dynamic Gesture Recognition Using MmWave Radar.IEEE Access10 (2022), 88692–88706. Securing Contrastive mmWave-based Human Activity Recognition against Adversarial Label Flipping WiSec ’24, May 27–30, 2024, Seoul, Republic of ...

    work page arXiv 2022

  5. [3]

    Joel Baptista, Vítor Santos, Filipe Silva, and Diogo Pinho. 2023. Domain Adap- tation with Contrastive Simultaneous Multi-Loss Training for Hand Gesture Recognition.Sensors23, 6 (2023). doi:10.3390/s23063332

    work page doi:10.3390/s23063332 2023

  6. [4]

    Ting Chen, Simon Kornblith, Mohammad Norouzi, and Geoffrey Hinton. 2020. A Simple Framework for Contrastive Learning of Visual Representations. arXiv:2002.05709 [cs.LG]

    arXiv 2020

  7. [5]

    Antonio Emanuele Cinà, Kathrin Grosse, Ambra Demontis, Sebastiano Vascon, Werner Zellinger, Bernhard A Moser, Alina Oprea, Battista Biggio, Marcello Pelillo, and Fabio Roli. 2023. Wild patterns reloaded: A survey of machine learning security against training data poisoning.Comput. Surveys55, 13s (2023), 1–39

    work page 2023

  8. [6]

    2008.Supervised Learning

    Pádraig Cunningham, Matthieu Cord, and Sarah Jane Delany. 2008.Supervised Learning. Springer Berlin Heidelberg, Berlin, Heidelberg, 21–49. doi:10.1007/978- 3-540-75171-7_2

    doi:10.1007/978- 2008

Show all 53 references
  1. [7]

    Yu Du, Yongkang Wong, Wenguang Jin, Wentao Wei, Yu Hu, Mohan Kankanhalli, and Weidong Geng. 2017. Semi-Supervised Learning for Surface EMG-Based Gesture Recognition. InProceedings of the 26th International Joint Conference on Artificial Intelligence(Melbourne, Australia)(IJCAI...

    2017

  2. [8]

    Tianyu Gu, Kang Liu, Brendan Dolan-Gavitt, and Siddharth Garg. 2019. Badnets: Evaluating backdooring attacks on deep neural networks.IEEE Access7 (2019), 47230–47244

    2019

  3. [9]

    Natabara Máté Gyöngyössy, Márk Domonkos, János Botzheim, and Péter Korondi

  4. [10]

    Harish Haresamudram, Irfan Essa, and Thomas Plötz. 2021. Contrastive Predictive Coding for Human Activity Recognition. (September 2021)

    2021

  5. [11]

    Cesar Iovescu and Sandeep Rao. 2017. The fundamentals of millimeter wave sensors.Texas Instruments(2017), 1–8

    2017

  6. [12]

    Prannay Khosla, Piotr Teterwak, Chen Wang, Aaron Sarna, Yonglong Tian, Phillip Isola, Aaron Maschinot, Ce Liu, and Dilip Krishnan. 2021. Supervised Contrastive Learning. arXiv:2004.11362 [cs.LG]

    2021 arXiv

  7. [13]

    Ang Li, Jiawei Li, Dianqi Han, Yan Zhang, Tao Li, and Yanchao Zhang. 2022. WearRF-CLA: Continuous Location Authentication with Wrist Wearables and UHF RFID. InACM AsiaCCS. Nagasaki

    2022

  8. [14]

    Chenning Li, Manni Liu, and Zhichao Cao. 2020. WiHF: Gesture and user recog- nition with WiFi.IEEE Transactions on Mobile Computing21, 2 (2020), 757–768

    2020

  9. [16]

    Yadong Li, Dongheng Zhang, Jinbo Chen, Jinwei Wan, Dong Zhang, Yang Hu, Qibin Sun, and Yan Chen. 2022. DI-Gesture: Domain-Independent and Real-Time Gesture Recognition with Millimeter-Wave Signals. InIEEE GLOBECOM. Rio de Janeiro, Brazil

    2022

  10. [17]

    Yadong Li, Dongheng Zhang, Jinbo Chen, Jinwei Wan, Dong Zhang, Yang Hu, Qibin Sun, and Yan Chen. 2022. Towards Domain-Independent and Real-Time Gesture Recognition Using Mmwave Signal.IEEE Transactions on Mobile Com- puting22 (December 2022), 7355–7369. Issue 12

    2022

  11. [18]

    Licsar and T

    A. Licsar and T. Sziranyi. 2002. Supervised training based hand gesture recog- nition system. In2002 International Conference on Pattern Recognition, Vol. 3. 999–1002 vol.3. doi:10.1109/ICPR.2002.1048206

    2002 arXiv

  12. [19]

    Haipeng Liu, Anfu Zhou, Zihe Dong, Yuyang Sun, Jiahe Zhang, Liang Liu, Huadong Ma, Jianhua Liu, and Ning Yang. 2022. M-Gesture: Person-Independent Real-Time In-Air Gesture Recognition Using Commodity Millimeter Wave Radar. IEEE Internet of Things Journal9, 5 (2022), 3397–3415....

    2022 doi

  13. [20]

    Jian Liu, Hongbo Liu, Yingying Chen, Yan Wang, and Chen Wang. 2019. Wireless sensing for human activity: A survey.IEEE Communications Surveys & Tutorials 22, 3 (2019), 1629–1645

    2019

  14. [21]

    Nitika Nigam, Tanima Dutta, and Hari Prabhat Gupta. 2020. Impact of Noisy Labels in Learning Techniques: A Survey. InAdvances in Data and Information Sciences, Mohan L. Kolhe, Shailesh Tiwari, Munesh C. Trivedi, and Krishn K. Mishra (Eds.). Springer Singapore, Singapore, 403–411

    2020

  15. [22]

    Yong Niu, Yong Li, Depeng Jin, Li Su, and Athanasios V Vasilakos. 2015. A survey of millimeter wave communications (mmWave) for 5G: opportunities and challenges.Wireless networks21 (2015), 2657–2676

    2015

  16. [24]

    Qifan Pu, Sidhant Gupta, Shyamnath Gollakota, and Shwetak Patel. 2013. Whole- home gesture recognition using wireless signals. InACM MobiCom. Miami, FL

    2013

  17. [25]

    Sandeep Rao. 2017. Introduction to mmWave sensing: FMCW radars.Texas Instruments (TI) mmWave Training Series(2017), 1–11

    2017

  18. [26]

    Abdur Shahid, Ahmed Imteaj, Peter Wu, Diane Igoche, and Tauhidul Alam

  19. [27]

    Abdur R Shahid, Ahmed Imteaj, Shahriar Badsha, and Md Zarif Hossain. 2023. Assessing wearable human activity recognition systems against data poison- ing attacks in differentially-private federated learning. InIEEE SMARTCOMP. Nashville, TN

    2023

  20. [28]

    Stephan Sigg, Shuyu Shi, Felix Buesching, Yusheng Ji, and Lars Wolf. 2013. Lever- aging RF-Channel Fluctuation for Activity Recognition: Active and Passive Sys- tems, Continuous and RSSI-Based Signal Features. InProceedings of International Conference on Advances in Mobile Com...

    2013

  21. [29]

    Ankit Singh, Omprakash Chakraborty, Ashutosh Varshney, Rameswar Panda, Rogerio Feris, Kate Saenko, and Abir Das. 2021. Semi-Supervised Action Recog- nition With Temporal Contrastive Learning. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (...

    2021

  22. [30]

    Hwanjun Song, Minseok Kim, Dongmin Park, Yooju Shin, and Jae-Gil Lee. 2022. Learning from noisy labels with deep neural networks: A survey.IEEE Transac- tions on Neural Networks and Learning Systems(2022)

    2022

  23. [31]

    Hwanjun Song, Minseok Kim, Dongmin Park, Yooju Shin, and Jae-Gil Lee. 2023. Learning From Noisy Labels With Deep Neural Networks: A Survey.IEEE Transactions on Neural Networks and Learning Systems34, 11 (2023), 8135–8153. doi:10.1109/TNNLS.2022.3152527

    2023

  24. [32]

    Ruiyuan Song, Dongheng Zhang, Zhi Wu, Cong Yu, Chunyang Xie, Shuai Yang, Yang Hu, and Yan Chen. 2022. Rf-url: unsupervised representation learning for rf sensing. InACM MobiCom. Sydney, Australia

    2022

  25. [33]

    Soumya, C

    A. Soumya, C. Krishna Mohan, and Linga Reddy Cenkeramaddi. 2023. Recent Ad- vances in mmWave-Radar-Based Sensing, Its Applications, and Machine Learning Techniques: A Review.Sensors23, 21 (2023). doi:10.3390/s23218901

    2023 doi

  26. [34]

    Zhiyi Tian, Lei Cui, Jie Liang, and Shui Yu. 2022. A comprehensive survey on poisoning attacks and countermeasures in machine learning.Comput. Surveys 55, 8 (2022), 1–35

    2022

  27. [35]

    Jie Wang, Xiao Zhang, Qinhua Gao, Hao Yue, and Hongyu Wang. 2017. Device- Free Wireless Localization and Activity Recognition: A Deep Learning Approach. IEEE Transactions on Vehicular Technology66, 7 (2017), 6258–6267. doi:10.1109/ TVT.2016.2635161

    2017

  28. [36]

    Wei Wang, Alex X Liu, Muhammad Shahzad, Kang Ling, and Sanglu Lu. 2017. Device-free human activity recognition using commercial WiFi devices.IEEE Journal on Selected Areas in Communications35, 5 (2017), 1118–1131

    2017

  29. [37]

    Songhua Wu, Xiaobo Xia, Tongliang Liu, Bo Han, Mingming Gong, Nannan Wang, Haifeng Liu, and Gang Niu. 2020. Class2Simi: A New Perspective on Learning with Label Noise.CoRRabs/2006.07831 (2020). arXiv:2006.07831 https: //arxiv.org/abs/2006.07831

    2020 arXiv

  30. [38]

    Huang Xiao, Battista Biggio, Blaine Nelson, Han Xiao, Claudia Eckert, and Fabio Roli. 2015. Support vector machines under adversarial label contamination. Neurocomputing160 (2015), 53–62

    2015

  31. [39]

    Han Xiao, Huang Xiao, and Claudia Eckert. 2012. Adversarial label flips attack on support vector machines. InECAI 2012. Montpellier, France

    2012

  32. [40]

    Tong Xiao, Tian Xia, Yi Yang, Chang Huang, and Xiaogang Wang. 2015. Learning From Massive Noisy Labeled Data for Image Classification. InProceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR)

    2015

  33. [41]

    Jiawei Xu, Ziqian Bi, Amit Singha, Tao Li, Yimin Chen, and Yanchao Zhang

  34. [42]

    Baiju Yan, Peng Wang, Lidong Du, Xianxiang Chen, Zhen Fang, and Yirong Wu

  35. [43]

    Xiaoyong Yuan, Pan He, Qile Zhu, and Xiaolin Li. 2019. Adversarial examples: Attacks and defenses for deep learning.IEEE transactions on neural networks and learning systems30, 9 (2019), 2805–2824

    2019

  36. [44]

    Chiyuan Zhang, Samy Bengio, Moritz Hardt, Benjamin Recht, and Oriol Vinyals

  37. [45]

    Dauphin, and David Lopez-Paz

    Hongyi Zhang, Moustapha Cisse, Yann N. Dauphin, and David Lopez-Paz. 2018. mixup: Beyond Empirical Risk Minimization. arXiv:1710.09412 [cs.LG]

    2018 arXiv

  38. [46]

    Jia Zhang, Rui Xi, Yuan He, Yimiao Sun, Xiuzhen Guo, Weiguo Wang, Xin Na, Yunhao Liu, Zhenguo Shi, and Tao Gu. 2023. A survey of mmWave-based human sensing: Technology, platforms and applications.IEEE Communications Surveys & Tutorials(2023)

    2023

  39. [47]

    doi:10.1016/j.eswa

    mmGesture: Semi-supervised gesture recognition system using mmWave radar.Expert Systems with Applications213 (2023), 119042. doi:10.1016/j.eswa. 2022.119042

    2023

  40. [53]

    Yongqiang Zhang, Lixin Peng, Guilei Ma, Menghua Man, and Shanghe Liu. 2022. Dynamic Gesture Recognition Model Based on Millimeter-Wave Radar With ResNet-18 and LSTM.Frontiers in Neurorobotics16 (06 2022). doi:10.3389/fnbot. 2022.903197

    2022

  41. [2016]

    arXiv:1611.03530 http://arxiv.org/abs/1611.03530

    Understanding deep learning requires rethinking generalization.CoRR abs/1611.03530 (2016). arXiv:1611.03530 http://arxiv.org/abs/1611.03530

    2016 arXiv

  42. [2019]

    In2019 IEEE Symposium Series on Computational Intelligence (SSCI)

    Supervised Learning with Small Training Set for Gesture Recognition by Spiking Neural Networks. In2019 IEEE Symposium Series on Computational Intelligence (SSCI). 2201–2206. doi:10.1109/SSCI44817.2019.9002720

    2019

  43. [2020]

    arXiv:2012.04462 https://arxiv.org/abs/2012.04462

    Multi-Objective Interpolation Training for Robustness to Label Noise.CoRR abs/2012.04462 (2020). arXiv:2012.04462 https://arxiv.org/abs/2012.04462

    2020 arXiv

  44. [2022]

    InIEEE SSCI

    Label Flipping Data Poisoning Attack Against Wearable Human Activity Recognition System. InIEEE SSCI. Singapore

  45. [2023]

    In2023 32nd International Conference on Computer Communications and Networks (ICCCN)

    mmLock: User Leaving Detection Against Data Theft via High-Quality mmWave Radar Imaging. In2023 32nd International Conference on Computer Communications and Networks (ICCCN). 1–10. doi:10.1109/ICCCN58024.2023. 10230151

    2023

Pith tools

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