Pith. sign in

REVIEW 5 major objections 3 minor 81 references

Silent Impact: Tracking Tennis Shots from the Passive Arm

T0 review · 5 major / 3 minor · reviewed 2026-08-06 · deepseek-v4-flash

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

Pith's one-line read The paper claims that a single IMU sensor on the passive arm can detect and classify six tennis shot types almost as accurately as the dominant arm, making a smartwatch already worn during play sufficient for shot analytics.

desk verdict The passive-arm idea is new and the experiments are mostly careful, but the detection F1 is conditional on dominant-arm-detectable shots, and the abstract overstates comparability. read the letter →

arxiv 2507.23215 v1 pith:JXXM3D42 submitted 2025-07-31 cs.HC

Junyong Park , Saelyne Yang , Sungho Jo This is my paper
classification cs.HC
keywords actionrecognitionmotionanalysisIMUtennissmartwatchshotclassificationdetectionwearablesensing
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

Silent Impact tries to show that tennis shot analysis does not need a sensor on the racket or the dominant arm: a smartwatch worn on the passive, non-dominant arm can detect when a shot happens and which of six shot types it is. The authors report 88.2% classification accuracy and an 86.0% detection F1 score from passive-arm data, within a few points of the dominant arm, and a user study in which players felt significantly less physical and mental burden. If true, recreational players could get automatic shot summaries from a device they already wear, with no extra gear and no change to their natural motion.

What carries the argument

The classification model decomposes the passive-arm IMU signal into low (0–4 Hz), medium (5–20 Hz), and high (above 20 Hz) frequency bands via the Fourier transform; each band feeds a temporal attention block that gates a one-dimensional convolutional backbone, letting the model exploit body-wide motion, arm swing, and impact vibration. Shot detection uses MS-TCN, a multi-stage temporal convolutional network for frame-wise action segmentation, followed by a refinement heuristic that merges consecutive 'true' frames into fixed-length shot windows.

What would settle it

Label a match session with a high-speed camera, marking every racket-ball contact by eye, then run the passive-arm detector against that ground truth and report detection F1 separately for volleys, drop shots, and half-volleys where the dominant-arm jerk is smallest; if F1 on those gentle shots falls well below 86%, the claim of comparable detection needs qualifying.

Watch

Extended reading notes

Core claim

Silent Impact establishes that the passive arm carries enough information about a tennis shot to serve as a standalone sensing site. With frequency-band attention, passive-arm classification reaches 88.2% accuracy versus 90.1% from the dominant arm, narrowing the gap of a standard backbone from 9.1 to 1.9 points; shot detection reaches 86.0% F1 versus 94.8% from the dominant arm, and a 10-person user study shows significantly lower mental and physical burden. The paper concludes that the passive arm is an effective, comfortable alternative for tennis shot analysis.

Load-bearing premise

The detection labels come from the dominant arm's acceleration peaks, so the passive-arm model is trained only to find shots forceful enough to jolt the dominant arm; if gentle shots produce no such peak, the reported detection F1 overstates real-world performance.

Editorial extensions

If this is right

  • A smartwatch on the passive arm can replace racket-attached or dominant-arm sensors for six-shot tennis analysis, with a 1.9-point accuracy cost in classification and an 8.8-point F1 cost in detection.
  • The frequency-band attention modules recover most of the passive-arm shortfall, improving accuracy by 6.8% over the unmodified backbone while barely changing dominant-arm results.
  • Shot detection from the passive arm requires learned models: simple threshold peak detection falls to 37.6 F1 on passive data, while the MS-TCN-based model raises it to 86.0.
  • Fine-tuning on just 10% of a player's own shots lifts classification accuracy to roughly 94%, suggesting a personalization path for real deployment.
  • The system runs as an end-to-end prototype on a commercial smartwatch with a mobile app, so the reported numbers correspond to a usable, not merely simulated, pipeline.

Reading between the lines

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

  • The reported detection F1 is measured against ground truth defined by dominant-arm acceleration peaks, so the comparison may undercount shots with soft impacts; a video-based label set could change the measured gap between arms.
  • The same subsidiary-motion principle should transfer to other racquet and stick sports such as badminton, golf, and table tennis, where the non-dominant limb moves with the body; the paper discusses this direction but does not demonstrate it.
  • Because sampling frequency at 30 Hz versus 120 Hz changes accuracy by less than 0.5%, the approach may also run on lower-cost wristbands rather than flagship smartwatches.
  • Combining passive-arm streams from both players in a match could infer point outcomes and rally structure without cameras, an application the paper mentions only as future collaborative use.
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 / 3 minor

Summary. Silent Impact proposes using a smartwatch on a player's passive (non-dominant) arm to detect and classify six tennis shots from IMU data, rather than requiring a sensor on the racket or dominant arm. The authors collect synchronized wrist IMU data from 20 recreational players (6000 shot windows for classification; 368 minutes of rally/match data from 10 players for detection), label shot events from dominant-arm acceleration peaks verified with video, and train a frequency-band attention CNN for shot classification and an MS-TCN model plus refinement heuristic for shot detection. They report 88.2±2.0% classification accuracy and 86.0% detection F1 with passive-arm data, versus 90.1±3.0% and 94.8% with dominant-arm data, and a 10-participant user study showing lower perceived burden. A smartwatch-to-cloud prototype is also presented.

Significance. If the reported performance transfers to real play, the paper makes a useful contribution to sports HCI: it demonstrates that a single commodity wrist-worn IMU on the passive arm can recover a substantial amount of shot information, and the public release of models and dataset supports reproducibility. Strengths include participant-disjoint 5-fold cross-validation, balanced shot sampling, two baselines (FCN and peak detection), and ablations on context, segment length, sensor modality, fine-tuning, and sampling rate. However, the detection ground truth is constructed from dominant-arm impact peaks, the headline F1 is frame-level, and the user-study statistics contain a test-design error; these issues currently limit the strength of the central "comparable to dominant arm" claim.

major comments (5)
  1. [§4.1.2, §5.2.2] Detection ground truth is not an exhaustive annotation of shots. §4.1.2 states that shot instances were found by thresholding the dominant arm's acceleration and that the manual video pass filters out incorrectly detected windows; it does not add shots the threshold missed. The paper itself notes in §4.2.2 that threshold methods "may fall short when detecting subtler shots, such as volleys or drop shots." Consequently, both training and evaluation for the 86.0% passive-arm F1 exclude gentle impacts that do not produce a dominant-arm jerk, so the reported F1 is an upper bound on passive-arm recall in real rallies. Please report detection performance separately for volley-like and other low-jerk events, or re-annotate the detection dataset exhaustively from video, or clearly re-frame the claim as detection of dominant-arm-detectable impacts.
  2. [§5.1, Table 2] The "detection F1" reported in Table 2 and in the abstract is a frame-level F1 for positive labels, as stated in §5.1, not a shot-level measure. Because a shot occupies a 180-frame window and the refinement heuristic merges predicted segments, frame-level F1 can remain high even when entire shots are missed or when partial overlaps count toward the score. The abstract's "detection F1 score of 86.0%" therefore may overstate shot-instance detection. Please report per-shot precision, recall, and F1 using a defined matching criterion (e.g., center-offset or IoU threshold).
  3. [§6.3.1, §6.3.2, Figure 9] The user study uses a within-subject design (§6.1), but the reported U statistics come from Mann-Whitney U tests, which assume two independent groups; the appropriate paired test is Wilcoxon signed-rank (or a paired t-test). In addition, §6.3.2 reports physical burden during play as U=85, p=0.07, which is not significant at the 0.05 level, while the Figure 9 caption claims that participants felt less mental and physical demand "by a statistically significant level, in both the setup process and playing tennis." This internal contradiction and the incorrect test make the user-comfort conclusions unreliable. Please re-analyze with paired tests, correct the caption, and report effect sizes.
  4. [Abstract, §1, §5.2] The abstract's "comparable to the dominant arm" is not supported for detection. Table 2 shows an 8.8-point F1 gap (86.0 vs 94.8), which is a substantial difference for shot detection, and no statistical test or confidence interval is given for either comparison. The classification gap of 1.9 points is more supportive, but grouping both metrics under "comparable" overstates the detection result. Please qualify the claim (e.g., "approaches dominant-arm performance" or state the gap explicitly) or provide statistical evidence for equivalence.
  5. [§4.3.2, §6] No accuracy evaluation is reported for the actual prototype hardware: all model numbers come from Xsens DOT sensors sampled at 120 Hz, while the Galaxy Watch 4 records at 100 Hz and is up-sampled to 120 Hz with linear interpolation (§4.3.2). The user study (§6) measures perceived burden but does not validate that the watch-based detection and classification preserve the reported accuracy. A small validation set recorded with the target smartwatch, or a clear statement that the prototype was not accuracy-tested, is needed to support the end-to-end prototype claim.
minor comments (3)
  1. [Appendix A.1.2] The phrase "increase in accuracy by 4 7%" appears to be missing a dash or range symbol; it should read "4–7%" or similar.
  2. [§5.3.2] The segment-length comparison reports accuracies of 79.4%, 85.9%, and 87.3% without error bars or significance tests; adding variance information would help the reader assess whether the differences are meaningful.
  3. [Figure 8] The timeline visualizations are dense and may be difficult to read in grayscale; adding explicit markers for missed detections and false positives would improve interpretability.

Circularity Check

0 steps flagged · score 1.0 of 10

No circular derivation: held-out passive-arm accuracy and F1 are genuine evaluations against externally verified labels; the single self-citation is not load-bearing.

full rationale

The paper's central claims are empirical measurements, not derivations from fitted constants. Shot classification accuracy (88.2±2.0% passive, 90.1±3.0% dominant) and detection F1 (86.0% vs 94.8%) are reported from 5-fold cross-validation with participant-disjoint folds (§5.1). No parameter is fitted to the test folds and then renamed a prediction. The shot labels are generated from dominant-arm acceleration peaks and then manually verified against video (§4.1.2), so the evaluation benchmark is external to the passive-arm model rather than defined by it. The passive-arm model's outputs are not algebraically equivalent to the dominant-arm threshold; the threshold is used only for annotation, not as a feature or prediction. The FCN and peak-detection baselines come from prior work [32]. The only self-citation is [55], used to support the generic observation that frame-wise action segmentation suffers from over-segmentation; this is background knowledge and does not carry the paper's thesis. A separate validity concern—that the threshold-based labels may omit gentle shots such as soft volleys—is a limitation of the dataset's completeness, not a circularity in the derivation chain.

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

The central claim rests on how shots are defined and labeled, on whether lab IMU data transfers to a commercial smartwatch, and on the model design choice that frequency bands separate informative motion components. These are reasonable domain assumptions, but each is load-bearing for the transfer from offline accuracy to a real product.

free parameters (5)
  • Shot window length = 1.5 seconds (180 frames at 120 Hz)
    Chosen in §4.1.1; ablation in §5.3.2 shows 1-second windows drop accuracy to 79.4%, so performance depends on this hand-selected value.
  • Detection class weight = 5:1 shot-to-non-shot
    Set by hand in §4.2.2 to compensate imbalance; directly shapes detection F1.
  • Frequency band cutoffs = 0-4 Hz, 5-20 Hz, above 20 Hz
    Selected in §4.2.1 from prior human-motion literature rather than from this dataset; the attention model's benefit depends on the bands isolating distinct signal components.
  • Refinement heuristic k = not reported
    §4.2.2 defines a run of k true frames as a candidate shot, but never states k's numeric value; it influences the final detection windows and cannot be reproduced exactly.
  • MS-TCN architecture size = 3 stages, 4 layers, hidden dimension 64
    Specified in §4.2.2; hand-picked capacity for frame-wise detection.
assumptions (4)
  • domain assumption Every true tennis shot produces a measurable dominant-arm acceleration peak that defines the shot moment.
    Used in §4.1.2 to auto-label detection ground truth; shots without a dominant-arm jerk, such as soft volleys, may be absent from the label set.
  • domain assumption Xsens DOT wrist IMU data is equivalent enough to Samsung Galaxy Watch 4 IMU data for trained models to transfer without accuracy re-validation.
    The prototype (§4.3.2) records at 100 Hz and up-samples to 120 Hz and aligns axes, but no experiment evaluates the deployed model on watch-collected data.
  • domain assumption The selected Fourier bands separate whole-body motion, arm swing, and impact vibrations.
    Model design in §4.2.1; justified only by references to prior human-motion work, not by data analysis in this paper.
  • standard math Participant-disjoint 5-fold cross-validation gives unbiased performance estimates.
    Standard evaluation protocol described in §5.1; valid assumption given random balanced assignment of participants to folds.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Silent Impact: Tracking Tennis Shots from the Passive Arm." pith.science (2026). https://pith.science/paper/JXXM3D42

@misc{pith2026250723215,
  author       = {Pith},
  title        = {Pith review of: Silent Impact: Tracking Tennis Shots from the Passive Arm},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/JXXM3D42}},
  note         = {Machine review of arXiv:2507.23215}
}
read the original abstract

Wearable technology has transformed sports analytics, offering new dimensions in enhancing player experience. Yet, many solutions involve cumbersome setups that inhibit natural motion. In tennis, existing products require sensors on the racket or dominant arm, causing distractions and discomfort. We propose Silent Impact, a novel and user-friendly system that analyzes tennis shots using a sensor placed on the passive arm. Collecting Inertial Measurement Unit sensor data from 20 recreational tennis players, we developed neural networks that exclusively utilize passive arm data to detect and classify six shots, achieving a classification accuracy of 88.2% and a detection F1 score of 86.0%, comparable to the dominant arm. These models were then incorporated into an end-to-end prototype, which records passive arm motion through a smartwatch and displays a summary of shots on a mobile app. User study (N=10) showed that participants felt less burdened physically and mentally using Silent Impact on the passive arm. Overall, our research establishes the passive arm as an effective, comfortable alternative for tennis shot analysis, advancing user-friendly sports analytics.

Figures

Figures reproduced from arXiv: 2507.23215 by the authors.

Figure 1
Figure 1. Unlike most analytical products that require sensors to be placed on the player’s dominant arm, Silent Impact analyzes [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. Comparison of the pose of the passive arm (circled [PITH_FULL_IMAGE:figures/full_fig_p002_2.png] view at source ↗
Figure 3
Figure 3. Participant demographics and tennis characteristics [PITH_FULL_IMAGE:figures/full_fig_p004_3.png] view at source ↗
Figures from the paper (6 more)
Figure 4
Figure 4. Figure 4: Sensor placement and orientation. For both the [PITH_FULL_IMAGE:figures/full_fig_p005_4.png]
Figure 5
Figure 5. Figure 5: Model architecture for shot classification. To capture maximal information from the motion of the passive arm, the [PITH_FULL_IMAGE:figures/full_fig_p006_5.png]
Figure 6
Figure 6. Figure 6: System architecture of Silent Impact. When the user initiates the app on their watch, the accelerometer and gyroscope [PITH_FULL_IMAGE:figures/full_fig_p007_6.png]
Figure 7
Figure 7. Figure 7: Comparison of the confusion matrices for shot clas [PITH_FULL_IMAGE:figures/full_fig_p008_7.png]
Figure 8
Figure 8. Figure 8: Shot detection results. Each timeline visualizes the predicted or ground truth labels of each frame within the sequence. [PITH_FULL_IMAGE:figures/full_fig_p009_8.png]
Figure 9
Figure 9. Figure 9: User rating on the mental and physical load as part [PITH_FULL_IMAGE:figures/full_fig_p010_9.png]

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

81 extracted references · 78 canonical work pages

  1. [1]

    Amin Ahmadi, David Rowlands, and Daniel Arthur James. 2009. Towards a wearable device for skill assessment and skill acquisition of a tennis player during the first serve. Sports Technology 2, 3-4 (2009), 129–136

    2009

  2. [2]

    Karan Ahuja, Sven Mayer, Mayank Goel, and Chris Harrison. 2021. Pose-on- the-go: Approximating user pose with smartphone sensor fusion and inverse kinematics. In Proceedings of the 2021 CHI Conference on Human Factors in Com- puting Systems. 1–12

    work page 2021

  3. [3]

    Gobinath Aroganam, Nadarajah Manivannan, and David Harrison. 2019. Review on wearable technology sensors used in consumer sport applications. Sensors 19, 9 (2019), 1983

    work page 2019

  4. [4]

    Babolat. 2023. Babolat PLAY. https://help.babolatplay.com/hc/en-us Accessed: August 27, 2023

    work page 2023

  5. [5]

    Driveline Baseball. 2023. PULSE Throw Workload Monitor . https://www. drivelinebaseball.com/product/pulse-throw/ Accessed: August 27, 2023

    work page 2023

  6. [6]

    David R Bassett, Lindsay P Toth, Samuel R LaMunion, and Scott E Crouter

    work page

  7. [7]

    Luis Benages Pardo, David Buldain Perez, and Carlos Orrite Urunuela. 2019. Detection of tennis activities with wearable sensors. Sensors 19, 22 (2019), 5004

    work page 2019

  8. [8]

    Sarnab Bhattacharya, Rebecca Adaimi, and Edison Thomaz. 2022. Leveraging sound and wrist motion to detect activities of daily living with commodity smartwatches. Proceedings of the ACM on Interactive, Mobile, Wearable and Ubiquitous Technologies 6, 2 (2022), 1–28

    work page 2022

Show all 81 references
  1. [9]

    Oli Bonar. 2023. Is the Qlipp the ultimate performance tennis sensor? https: //kitradar.com/blogs/reviews/is-the-qlipp-the-ultimate-performance-tennis- sensor#:~:text=RECREATIONAL%20AND%20ELITE%20PLAYER%20REVIEW Accessed: August 27, 2023

    2023

  2. [10]

    Jernej Borovinšek. 2011. Polona Hercog at the Fed Cup vs Germany in Maribor (3).jpg. http://j3rn3jphoto.blogspot.com/2011/02/fed-cup-slovenija-vs-nemcija. html. Licensed under the Creative Commons Attribution-Share Alike 2.5 Slovenia license

    2011

  3. [11]

    Howard Brody. 2000. Player sensitivity to the moments of inertia of a tennis racket. Sports Engineering 3, 2 (2000), 145–148

    2000

  4. [12]

    Seonjeong Byun, Hyang Jun Lee, Ji Won Han, Jun Sung Kim, Euna Choi, and Ki Woong Kim. 2019. Walking-speed estimation using a single inertial measure- ment unit for the older adults. PLoS one 14, 12 (2019), e0227075

    2019

  5. [13]

    Claudia Carello, Steve Thuot, Krista L Anderson, and MT Turvey. 1999. Perceiving the sweet spot. Perception 28, 3 (1999), 307–320

    1999

  6. [14]

    Catapult. 2023. Catapult. https://www.catapult.com/ Accessed: August 27, 2023

    2023

  7. [15]

    Liqiong Chang, Jiaqi Lu, Ju Wang, Xiaojiang Chen, Dingyi Fang, Zhanyong Tang, Petteri Nurmi, and Zheng Wang. 2018. SleepGuard: Capturing rich sleep information using smartwatch sensing data.Proceedings of the ACM on Interactive, Mobile, Wearable and Ubiquitous Technologies 2, ...

    2018

  8. [16]

    Dea Cizmic, Dominik Hoelbling, René Baranyi, Roland Breiteneder, and Thomas Grechenig. 2023. Smart Boxing Glove “RD 𝛼”: IMU Combined with Force Sensor for Highly Accurate Technique and Target Recognition Using Machine Learning. Applied Sciences 13, 16 (2023), 9073

    2023

  9. [18]

    Steve Collis. 2010. Philipp Kohlschreiber (4308345979).jpg. https://www.flickr. com/photos/stevecollis/4308345979/. Licensed under the Creative Commons Attribution 2.0 Generic license

    2010

  10. [19]

    Zepp Health Corporation. 2023. ZEPP. https://www.zepp.com/ Accessed: August 27, 2023

    2023

  11. [20]

    Božidara Cvetkovic, Vid Drobnic, and Mitja Luštrek. 2017. Recognizing hand- specific activities with a smartwatch placed on dominant or non-dominant wrist. In Proceedings of the 20th International Conference Information Society IS . 75–78

    2017

  12. [21]

    Keith Davids, Simon J Bennett, and Sam Beak. 2004. Sensitivity of children and adults to haptic information in wielding tennis rackets. In Interceptive actions in sport. Routledge, 195–211

    2004

  13. [22]

    Nathan DeVrio, Vimal Mollyn, and Chris Harrison. 2023. SmartPoser: Arm Pose Estimation with a Smartphone and Smartwatch Using UWB and IMU Data. In Proceedings of the 36th Annual ACM Symposium on User Interface Software and Technology. 1–11

    2023

  14. [23]

    Olivier Dieu, Jacques Mikulovic, Paul S Fardy, Gilles Bui-Xuan, Laurent Béghin, and Jérémy Vanhelst. 2017. Physical activity using wrist-worn accelerometers: comparison of dominant and non-dominant wrist. Clinical physiology and func- tional imaging 37, 5 (2017), 525–529

    2017

  15. [24]

    Chris Eason. 2009. Soderling Roland Garros 2009 4.jpg. https://commons. wikimedia.org/wiki/File:Soderling_Roland_Garros_2009_4.jpg. Licensed un- der the Creative Commons Attribution 2.0 Generic license

    2009

  16. [25]

    Dan Eisenhardt, Aidan Kits, Pascal Madeleine, Afshin Samani, David C Clarke, and Mathias Kristiansen. 2023. Augmented-reality swim goggles accurately and reliably measure swim performance metrics in recreational swimmers. Frontiers in Sports and Active Living 5 (2023), 1188102

    2023

  17. [26]

    Don Samitha Elvitigala, Armağan Karahanoğlu, Andrii Matviienko, Laia Turmo Vidal, Dees Postma, Michael D Jones, Maria F Montoya, Daniel Har- rison, Lars Elbæk, Florian Daiber, et al. 2024. Grand Challenges in SportsHCI. In Proceedings of the CHI Conference on Human Factors in ...

    2024

  18. [27]

    Nazli FarajiDavar, Teófilo De Campos, Josef Kittler, and Fei Yan. 2011. Trans- ductive transfer learning for action recognition in tennis games. In 2011 IEEE International Conference on Computer Vision Workshops (ICCV Workshops) . IEEE, 1548–1553

    2011

  19. [28]

    Yazan Abu Farha and Jurgen Gall. 2019. Ms-tcn: Multi-stage temporal convolu- tional network for action segmentation. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition . 3575–3584

    2019

  20. [29]

    João J Ferreira, Cristina I Fernandes, Hussain G Rammal, and Pedro M Veiga

  21. [30]

    Fitbit. 2023. Fitbit. https://www.fitbit.com/global/us/home Accessed: August 27, 2023

    2023

  22. [31]

    Carine06 from UK. 2011. Juan Martin del Potro (8610260814).jpg. https://www. flickr.com/photos/15686285@N06/8610260814/. Licensed under the Creative Commons Attribution-Share Alike 2.0 Generic license

    2011

  23. [32]

    Andreas Ganser, Bernhard Hollaus, and Sebastian Stabinger. 2021. Classification of tennis shots with a neural network approach. Sensors 21, 17 (2021), 5703

    2021

  24. [33]

    Garmin. 2023. Garmin. https://www.garmin.com/en-US/ Accessed: August 27, 2023

    2023

  25. [34]

    Sofia Gourgari, Georgios Goudelis, Konstantinos Karpouzis, and Stefanos Kollias

  26. [35]

    Fasih Haider, Fahim Salim, Vahid Naghashi, Sena Busra Yengec Tasdemir, Izem Tengiz, Kubra Cengiz, Dees Postma, Robby van Delden, Dennis Reidsma, Bert- Jan van Beijnum, et al. 2019. Evaluation of dominant and non-dominant hand movements for volleyball action modelling. In Adjun...

    2019

  27. [36]

    HEAD. 2023. Head Tennis Sensor. https://www.head.com/en_US/sensor Accessed: August 27, 2023

    2023

  28. [37]

    Hexoskin. 2023. Hexoskin. https://www.hexoskin.com/ Accessed: August 27, 2023

    2023

  29. [38]

    Mangolytics Inc. 2023. SwingVision. https://swing.tennis/ Accessed: August 27, 2023

    2023

  30. [39]

    CCS Insight. 2022. Smart wearable shipments forecast worldwide from 2016 to 2026 (in millions). Statista. https://www.statista.com/statistics/878144/ worldwide-smart-wristwear-shipments-forecast/ [Graph]

    2022

  31. [40]

    Tianjian Ji and Aikaterini Pachi. 2005. Frequency and velocity of people walking. Struct. Eng 84, 3 (2005), 36–40

    2005

  32. [41]

    Hong Jia, Jiawei Hu, and Wen Hu. 2021. SwingNet: Ubiquitous Fine-Grained Swing Tracking Framework via Stochastic Neural Architecture Search and Ad- versarial Learning. Proceedings of the ACM on Interactive, Mobile, Wearable and Ubiquitous Technologies 5, 3 (2021), 1–21

    2021

  33. [42]

    Rinat Khusainov, Djamel Azzi, Ifeyinwa E Achumba, and Sebastian D Bersch. 2013. Real-time human ambulation, activity, and physiological monitoring: Taxonomy of issues, techniques, applications, challenges and limitations. Sensors 13, 10 (2013), 12852–12902

    2013

  34. [43]

    Kickstarter. 2023. Smash. https://www.kickstarter.com/projects/1123712033/ smash-the-game-changing-tennis-wearable Accessed: August 27, 2023

    2023

  35. [44]

    Marko Kos and Iztok Kramberger. 2018. Tennis stroke consistency analysis using miniature wearable IMU. In2018 25th International Conference on Systems, Signals and Image Processing (IWSSIP) . IEEE, 1–4

    2018

  36. [45]

    Gierad Laput and Chris Harrison. 2019. Sensing fine-grained hand activity with smartwatches. In Proceedings of the 2019 CHI Conference on Human Factors in Computing Systems. 1–13

    2019

  37. [46]

    Athas, Andreas Edgar SCHOBEL, Shawn R

    Gierad Laput, Jared LeVan Zerbe, William C. Athas, Andreas Edgar SCHOBEL, Shawn R. Scully, Brian H. Tsang, Kevin Lynch, Charles MAALOUF, and Shiwen ZHAO. 2020. Methods and apparatus for detecting individual health related events. https://patents.google.com/patent/US20210063434...

    2020

  38. [47]

    Hong Li, Shishir Chawla, Richard Li, Sumeet Jain, Gregory D Abowd, Thad Starner, Cheng Zhang, and Thomas Plötz. 2018. Wristwash: towards automatic handwashing assessment using a wrist-worn device. In Proceedings of the 2018 ACM international symposium on wearable computers . 132–139

    2018

  39. [48]

    Guillaume Lopez, Shohei Abe, Kengo Hashimoto, and Anna Yokokubo. 2019. On-site personal sport skill improvement support using only a smartwatch. In 2019 IEEE International Conference on Pervasive Computing and Communications Workshops (PerCom Workshops). IEEE, 158–164

    2019

  40. [49]

    Mark Mirtchouk, Drew Lustig, Alexandra Smith, Ivan Ching, Min Zheng, and Samantha Kleinberg. 2017. Recognizing eating from body-worn sensors: Com- bining free-living and laboratory data. Proceedings of the ACM on Interactive, Mobile, Wearable and Ubiquitous Technologies 1, 3 (...

    2017

  41. [50]

    Diganta Misra. 2019. Mish: A self regularized non-monotonic activation function. arXiv preprint arXiv:1908.08681 (2019). UIST ’24, October 13–16, 2024, Pittsburgh, PA, USA Junyong Park, Saelyne Yang, and Sungho Jo

    2019 arXiv

  42. [51]

    Vimal Mollyn, Riku Arakawa, Mayank Goel, Chris Harrison, and Karan Ahuja

  43. [52]

    Moov. 2023. Moov HR Sweat. https://welcome.moov.cc/moovhr Accessed: August 27, 2023

    2023

  44. [53]

    Movella. 2023. Xsens DOT User Manual. https://www.movella.com/hubfs/Xsens% 20DOT%20User%20Manual.pdf Accessed: August 27, 2023

    2023

  45. [54]

    Octav Opaschi and Radu-Daniel Vatavu. 2020. Uncovering practical security and privacy threats for connected glasses with embedded video cameras. Proceedings of the ACM on Interactive, Mobile, Wearable and Ubiquitous Technologies4, 4 (2020), 1–26

    2020

  46. [55]

    Junyong Park, Daekyum Kim, Sejoon Huh, and Sungho Jo. 2022. Maximization and restoration: Action segmentation through dilation passing and temporal reconstruction. Pattern Recognition 129 (2022), 108764

    2022

  47. [56]

    Weiping Pei, Jun Wang, Xubin Xu, Zhengwei Wu, and Xiaorong Du. 2017. An embedded 6-axis sensor based recognition for tennis stroke. In 2017 IEEE Interna- tional Conference on Consumer Electronics (ICCE) . IEEE, 55–58

    2017

  48. [57]

    PIQ. 2023. PIQ User Guide . http://cdn.cnetcontent.com/38/43/3843f3aa-6cc7- 467c-bced-99227662f34e.pdf Accessed: August 27, 2023

    2023

  49. [58]

    QLIPP. 2023. QLIPP Sensor. https://www.qlipp.com/20160303QLIPPManual.pdf Accessed: August 27, 2023

    2023

  50. [59]

    Amazon Web Services. 2023. Amazon S3. https://aws.amazon.com/pm/serv-s3/ Accessed: August 27, 2023

    2023

  51. [60]

    Sheng Shen, He Wang, and Romit Roy Choudhury. 2016. I am a smartwatch and i can track my user’s arm. In Proceedings of the 14th annual international conference on Mobile systems, applications, and services . 85–96

    2016

  52. [61]

    Maria Skublewska-Paszkowska and Pawel Powroznik. 2023. Temporal Pattern Attention for Multivariate Time Series of Tennis Strokes Classification. Sensors 23, 5 (2023), 2422

    2023

  53. [62]

    Maria Skublewska-Paszkowska, Pawel Powroznik, and Edyta Lukasik. 2020. Learning three dimensional tennis shots using graph convolutional networks. Sensors 20, 21 (2020), 6094

    2020

  54. [63]

    Sony. 2023. Sony Smart Tennis Sensor. https://developer.sony.com/smart-tennis- sensor-sdk Accessed: August 27, 2023

    2023

  55. [64]

    USA Steven Pisano from Brooklyn, NY. 2014. 2014 US Open (Tennis) - Tournament - Flavia Pennetta (14908375337).jpg. https://www.flickr.com/photos/stevenpisano/ 14908375337/. Licensed under the Creative Commons Attribution 2.0 Generic license

    2014

  56. [65]

    Jason D Stone, Lauren E Rentz, Jillian Forsey, Jad Ramadan, Rachel R Markwald, Victor S Finomore, Scott M Galster, Ali Rezai, and Joshua A Hagen. 2020. Evalua- tions of commercial sleep technologies for objective monitoring during routine sleeping conditions. Nature and scienc...

    2020

  57. [66]

    Sara Taghavi, Fardjad Davari, Hadi Tabatabaee Malazi, and Ahmad Ali Abin

  58. [67]

    Franck Tchuente, Natalie Baddour, and Edward D Lemaire. 2020. Classification of aggressive movements using smartwatches. Sensors 20, 21 (2020), 6377

    2020

  59. [68]

    TuringSense

    Inc. TuringSense. 2023. PIVOT. https://www.turingsense.com/ Accessed: August 27, 2023

    2023

  60. [69]

    Dhruv Verma, Sejal Bhalla, Dhruv Sahnan, Jainendra Shukla, and Aman Parnami

  61. [70]

    Tran Huy Vu, Archan Misra, Quentin Roy, Kenny Choo Tsu Wei, and Youngki Lee. 2018. Smartwatch-based early gesture detection 8 trajectory tracking for interactive gesture-driven applications. Proceedings of the ACM on Interactive, Mobile, Wearable and Ubiquitous Technologies 2,...

    2018

  62. [71]

    Stanley, and Karthik Jayaraman Raghuram

    Di Wu, James Ochs, Paige N. Stanley, and Karthik Jayaraman Raghuram. 2020. Estimating caloric expenditure using heart rate model specific to motion class. https://patents.google.com/patent/US11638556B2 Assigned to Apple Inc

    2020

  63. [72]

    Mingyue Wu, Mengjiao Fan, Yang Hu, Ran Wang, Yufan Wang, Yanchun Li, Shengyuan Wu, and Guowen Xia. 2022. A real-time tennis level evaluation and strokes classification system based on the Internet of Things. Internet of Things 17 (2022), 100494

    2022

  64. [73]

    Min Wu and Jake Luo. 2019. Wearable technology applications in healthcare: a literature review. Online J. Nurs. Inform 23, 3 (2019)

    2019

  65. [74]

    WEARABLE X. 2023. Wearable X. https://www.wearablex.com/xxxxx Accessed: August 27, 2023

    2023

  66. [75]

    Proceed- ings of the ACM on Interactive, Mobile, Wearable and Ubiquitous Technologies 5, 3 (2021), 1–28

    Expressear: Sensing fine-grained facial expressions with earables. Proceed- ings of the ACM on Interactive, Mobile, Wearable and Ubiquitous Technologies 5, 3 (2021), 1–28

    2021

  67. [76]

    forehand stroke

    Zhendong Zhuang and Yang Xue. 2019. Sport-related human activity detection and recognition using a smartwatch. Sensors 19, 22 (2019), 5001. Silent Impact: Tracking Tennis Shots from the Passive Arm UIST ’24, October 13–16, 2024, Pittsburgh, PA, USA A APPENDIX A.1 Additional Ex...

    2019

  68. [81]

    Disheng Yang, Jian Tang, Yang Huang, Chao Xu, Jinyang Li, Liang Hu, Guobin Shen, Chieh-Jan Mike Liang, and Hengchang Liu. 2017. TennisMaster: An IMU- based online serve performance evaluation system. In Proceedings of the 8th augmented human international conference . 1–8

    2017

  69. [2013]

    In Proceed- ings of the IEEE conference on computer vision and pattern recognition workshops

    Thetis: Three dimensional tennis shots a human action dataset. In Proceed- ings of the IEEE conference on computer vision and pattern recognition workshops . 676–681

  70. [2017]

    Sports Medicine 47 (2017), 1303–1315

    Step counting: a review of measurement considerations and health-related applications. Sports Medicine 47 (2017), 1303–1315

    2017

  71. [2019]

    In 2019 9th International Conference on Computer and Knowledge Engineering (ICCKE)

    Tennis stroke detection using inertial data of a smartwatch. In 2019 9th International Conference on Computer and Knowledge Engineering (ICCKE) . IEEE, 466–474

    2019

  72. [2021]

    Computers in Human Behavior 118 (2021), 106710

    Wearable technology and consumer interaction: A systematic review and research agenda. Computers in Human Behavior 118 (2021), 106710

    2021

  73. [2023]

    In Proceedings of the 2023 CHI Conference on Human Factors in Computing Systems

    IMUPoser: Full-Body Pose Estimation using IMUs in Phones, Watches, and Earbuds. In Proceedings of the 2023 CHI Conference on Human Factors in Computing Systems. 1–12

    2023

Pith tools

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