Pith. sign in

REVIEW 3 major objections 6 minor 1 cited by

Boosting Automatic Exercise Evaluation Through Musculoskeletal Simulation-Based IMU Data Augmentation

T0 review · 3 major / 6 minor · reviewed 2026-08-07 · deepseek-v4-flash

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

Pith's one-line read This paper argues that musculoskeletal simulation combined with trajectory modification and automatic threshold-based labeling can generate realistic IMU movement data and thereby improve exercise-quality classification, generalization to…

desk verdict A promising augmentation pipeline with honest FDE results, but the auto-labeling of simulated data is never independently validated, so the general 'boosting' claim is only conditionally supported. read the letter →

arxiv 2505.24415 v2 pith:LOSOULAH submitted 2025-05-30 cs.LG cs.AI

Andreas Spilz , Heiko Oppel , Michael Munz This is my paper
classification cs.LGcs.AI
keywords IMUdataaugmentationmusculoskeletalsimulationautomaticexerciseevaluationinversekinematicsphysiotherapydeeplearningFunctionalMovementScreeningpatient-specificfine-tuning
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 sets out to solve a data problem that blocks automated physiotherapy assessment: IMU recordings of exercises are scarce, imbalanced across quality classes, and hard to label. It proposes to generate new training examples by taking recorded movement trajectories, systematically shifting their starting posture and range of motion, and running them through a musculoskeletal inverse-kinematics simulation so the resulting movements respect anatomical joint constraints. Because the simulation changes which quality label a movement deserves, the method adds an automatic labeling step: kinematic outputs are checked against rules derived from the exercise's scoring criteria, with thresholds tuned to match expert labels on real data. Across three exercise datasets, the augmented examples resemble real data and, when added to training, improve classification, generalization to unseen subjects, and fine-tuning from two real repetitions.

What carries the argument

The mechanism has three linked parts. First, each recorded IMU orientation trajectory is converted to Euler angles and modified by $e_{s,\mathrm{aug}} = (e_s[t] - e_s[0]) \odot \alpha + \beta$, where the offset $\beta$ encodes initial posture and the scaling $\alpha$ encodes a target range of motion; both are sampled from class-, exercise-, and segment-specific multivariate normal distributions. Second, an inverse-kinematics solve against a full-body musculoskeletal model enforces anatomical constraints so the edited trajectory remains plausible. Third, automatic labeling evaluates kinematic parameters against per-exercise criteria, with decision thresholds found by random search over 10 million combinations to maximize geometric-mean F1 against physiotherapist labels on real data; generated examples are kept only when the automatic label agrees with the class sampled by the augmentation distribution.

What would settle it

Have independent physiotherapists blind-label a random sample of the augmented repetitions and compare their labels to the automatic labels per class and per dataset; substantial disagreement, especially on Hurdle Step class 1 (represented by only eight real examples), would falsify the claim that augmentation provides reliably labeled training data.

Watch

Extended reading notes

Core claim

On the paper's own terms, the central claim is that a non-label-preserving augmentation pipeline—trajectory modification plus musculoskeletal simulation plus automatic labeling—can manufacture realistic IMU orientation sequences for physiotherapeutic exercises without relying on motion capture, video, or pose data. The evidence is that augmented-only training nearly matches real-only training on the Foot Drop exercise (macro F1 0.933 versus 0.952), that combining real and augmented data raises leave-one-subject-out performance for most participants, and that patient-specific fine-tuning with two real examples reaches near-perfect scores when augmented variants supply the missing classes. The paper also reports limits: on Deep Squat and Hurdle Step, where expert labels are ambiguous, gains shrink and augmented-only training degrades, indicating that label ambiguity rather than biomechanical fidelity is the binding constraint.

Load-bearing premise

The load-bearing premise is that the automatic-labeling thresholds, fitted to reproduce physiotherapist labels on real recordings, remain correct when applied to simulated augmented movements; if they misfit, the generated training targets are unreliable and the reported gains are built on mislabeled data.

Editorial extensions

If this is right

  • Training sets for IMU-based exercise classifiers can be enlarged without new recording campaigns, including targeted expansion of minority classes.
  • Manual expert labeling of generated movements can be replaced by the threshold-based automatic labeling, reducing cost and enabling reproducible labels.
  • Pretrained networks can be personalized to a new patient from as few as two real repetitions when augmented variants of those repetitions supply the other classes.
  • The size of the gains depends on dataset properties: balanced, objectively labeled exercises benefit most, while subjectively labeled exercises are limited by label ambiguity.
  • Augmented data appear to cover a broader range of movement variance than the small real datasets, so combining real and augmented training is preferable to either alone.

Reading between the lines

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

  • If the threshold-generalization premise holds, the same pipeline could be adapted to any exercise whose scoring criteria can be expressed as kinematic rules, turning clinical protocols into generators of labeled training data.
  • The subject-specific clustering visible in the dimensionality-reduction analysis suggests a further use the paper does not develop: augmented examples of a particular patient could serve as a data-driven prior for domain adaptation or home-monitoring anomaly detection.
  • One testable extension is to replace the hard threshold rules with a probabilistic labeler that outputs a confidence per class, which would let the augmentation filter retain ambiguous examples instead of discarding them and might soften the label-ambiguity bottleneck on Deep Squat and Hurdle Step.
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

3 major / 6 minor

Summary. The paper proposes a data augmentation method for IMU-based physiotherapy exercise classification. Starting from real IMU orientation trajectories, the method applies per-segment Euler-angle offsets and range scalings whose parameters are sampled from class- and segment-specific multivariate normal distributions, then projects the modified trajectories through OpenSim inverse kinematics to enforce biomechanical plausibility. Labels for the augmented examples are assigned by an automatic ruleset whose decision thresholds are optimized by random search over 10 million combinations to maximize the geometric mean of class-wise F1-scores against physiotherapist labels on the real data. The authors evaluate the augmented data with t-SNE visualizations, cross-dataset transfer experiments (TRTR, TATR, TRTA), leave-one-subject-out cross-validation, and patient-specific fine-tuning, using three datasets (Foot Drop exercise, Deep Squat, Hurdle Step) with increasing class imbalance and label ambiguity. The paper reports improvements in LOSOCV and fine-tuning scenarios when augmented data are added, alongside notable degradations in the transfer-only scenarios.

Significance. If the label-transfer premise holds, the method is a practically valuable contribution: it produces biomechanically plausible IMU trajectories without requiring MoCap or video data, it does not need a GPU, and it offers a route to enlarge minority classes in physiotherapy exercise datasets. The paper is unusually honest in reporting the settings where augmentation hurts (TATR drops for DS and HS, TRTA failures, subject 105 degradation), and the experimental design is clearly described with a sensible separation of train/test folds. The main weakness is that the automatic labeling of augmented examples rests on thresholds fitted to real data with no independent check on simulated kinematics; this is the paper's load-bearing assumption and it is not validated.

major comments (3)
  1. [Methods, Automatic labeling; Methods, Generation of augmented examples] The automatic labeling thresholds are selected by random search over 10 million combinations to maximize GMF1 against physiotherapist labels on the real dataset (Table 1), and the same thresholds are subsequently applied to the augmented examples. The pipeline retains only augmented examples whose auto-label matches the class distribution from which the augmentation parameters were sampled. This makes the augmented labels a product of the same fitted ruleset by construction, and the optimization target is real-data agreement only; there is no held-out validation of threshold transfer to simulated kinematics. The paper's own discussion states that 'labeling of the augmented examples is significantly clearer than that of the real data,' which is an assumption, not a measurement. Because the downstream classification and fine-tuning gains rest on the correctness of these labels, this is a load-bearing gap. I recommend validating the auto-label on a subset of augmented examples with expert physiotherapist labels, or fitting thresholds on a leave-one-subject-out basis and reporting the resulting variability.
  2. [Results, Table 2 and discussion] The abstract claims that augmented variants 'contribute to gains in classification accuracy, generalization to unseen subjects, and patient-specific fine-tuning from few examples,' but the controlled transfer experiments in Table 2 show substantial degradation. In TATR, the macro F1 drops from 0.92 to 0.81 for DS and from 0.79 to 0.51 for HS, and in TRTA real-trained models fail on augmented data, with HS class 1 at F1 = 0.00. These results do not support the strong claim that augmented data 'closely resemble real-world data' in a distributional sense. The paper hypothesizes that the augmented data have larger variance than the real data, but this is not quantified. Please either temper the abstract's claim or provide an analysis that reconciles the strong transfer failures with the LOSOCV gains, for example by reporting per-class feature-space distances or a calibrated metric of distribution shift.
  3. [Results, leave-one-subject-out cross-validation (Figure 6); Discussion] The LOSOCV results are the main positive evidence for the method, but they are presented without subject-level error bars or a statistical significance test across folds, and the reported gains are not uniform: subject 105 in the HS dataset shows a pronounced degradation, and several subjects show little or negative change. Given that the paper frames the method as improving generalization to unseen subjects, the single-subject degradation and the lack of confidence intervals should be addressed, e.g., by reporting per-subject confidence intervals, a paired test across subjects, or an analysis of why subject 105 is an outlier in terms of class distribution or ambiguity.
minor comments (6)
  1. [Abstract] Typo: 'augmented variants closely resembles real-world data' should be 'closely resemble.'
  2. [Methods, Neural network architecture and training details] Typo: 'ouput layer' should be 'output layer.'
  3. [Methods, Patient specific finetuning] The learning rate schedule is cited as '[QUELLE]', which is a placeholder and must be replaced with an actual reference.
  4. [Results, Figure 1 caption] The caption contains garbled text ('FMS 1 / FDE CE FMS 2 / FDE FD FMS 3 / FDE WSSL') that appears to be a legend leftover; please clean it up.
  5. [Discussion] The sentence 'A plausible explanation for the observed discrepancies is possibly the different frequencies' is redundant; consider tightening.
  6. [Data available] The availability statement says data are available 'on reasonable request' without mentioning source code or the augmentation pipeline; given the methodological focus, providing code or a detailed configuration would strengthen reproducibility.

Circularity Check

3 steps flagged · score 6.0 of 10

Automatic-label thresholds are fit to the full real dataset and to the same GMF1 values reported as validation; these thresholds then both label and filter augmented examples, so the augmented-label 'reliability' and the LOSO/fine-tuning gains are partly constructed from the very real labels they are supposed to help predict.

  1. fitted input called prediction [Methods, 'Automatic labeling' (random-search threshold optimization); Table 1.]
    "To optimize the threshold values, we employ a random search strategy... We evaluated a total of 10 million threshold combinations per exercise, selecting the set yielding the highest GMF1 score for subsequent classification of augmented data. The achieved GMF1 scores for the three used exercises are denoted in table 1. This procedure ensures that the classification framework aligns as closely as possible with expert-based assessment, enabling reliable labeling of the augmented repetitions."

    The Table 1 GMF1 values are the optimization objective evaluated on the same real physiotherapist-labeled data that produced the thresholds; they are training scores, not held-out measures. These same thresholds then label every augmented example. The paper's inference that augmented repetitions are 'reliably' labeled therefore rests on the fit itself: no independent labels for augmented data are ever obtained. This is a fitted input presented as a validated capability.

  2. self definitional [Methods, 'Generation of augmented examples' (selection filter).]
    "To ensure only unambiguous examples were created, a preliminary selection process was implemented. Only examples whose automatic labeling matched the class assigned by the selected distribution were retained for further analysis."

    The criterion for an 'unambiguous' augmented example is agreement with the automatic labeling system, which is the same fitted ruleset that assigned the label. This selection cannot provide evidence about label quality; it simply enforces internal consistency between the sampled class and the ruleset's output. The augmented training labels are therefore constructed by definition from the fitted ruleset, rather than independently verified.

1 more flagged steps
  1. other [Methods 'Automatic labeling' (threshold fit on full data) + LOSOCV experiments.]
    "the real dataset is then processed through the automated classification framework, and the alignment between the automated labels and physiotherapist-assigned labels is assessed using the geometric mean across the F1-scores of the n individual classes GMF1... We evaluated a total of 10 million threshold combinations per exercise, selecting the set yielding the highest GMF1 score for subsequent classification of augmented data. ... In each LOSOCV iteration, one subject was entirely excluded from the training and validation sets and included solely in the test set."

    The threshold fit is performed on the full real dataset before the LOSOCV splits. Therefore the held-out subject's physiotherapist labels are part of the objective used to select the thresholds that later label augmented training examples for that subject's fold. The augmented training targets thus encode test-label information, so the reported cross-subject generalization and fine-tuning gains are inflated by construction rather than being independent measurements.

full rationale

The paper's central novelty is the musculoskeletal simulation pipeline and the claim that its automatic labeling is reliable and that augmented data improve cross-subject generalization and few-shot fine-tuning. The simulation and inverse-kinematics parts are self-contained and not circular: the Euler-angle modifications are compared against real data, and the musculoskeletal constraints are externally defined. However, the labeling step is load-bearing and circular in three connected places. First, the automatic-label thresholds are chosen by random search to maximize GMF1 on the full real dataset, and the same GMF1 values are then reported (Table 1) as evidence that the automatic labels are reliable for augmented data; this is a fitted-input-called-prediction pattern, since the validation score is the optimization objective on the fitting data. Second, augmented examples are kept only when their automatic label agrees with the class sampled from the distribution, so 'unambiguous' is defined as agreement with the very ruleset that produced the label; the label is constructed, not independently checked. Third, because the threshold fit uses all real data before the LOSOCV splits are made, the held-out subject's physiotherapist labels leak into the thresholds that label augmented training samples for that subject's fold, compromising the headline generalization and fine-tuning results. The TATR and t-SNE evidence retains some independent content, but the paper's strongest claims about unseen-subject generalization and patient-specific fine-tuning are partially forced by this leak, so a score of 6 is appropriate.

Assumptions & free parameters 2 free parameters · 3 assumptions · 0 invented entities

The central claim rests on two fitted components: class-specific augmentation statistics and labeling thresholds, both derived from the real training data. The paper does not introduce new physical entities. The main imported assumptions are the validity of the OpenSim musculoskeletal model and the FMS/FDE evaluation criteria.

free parameters (2)
  • Augmentation distribution statistics (means and stds of offset β and range target δ per segment, Euler component… = Not reported numerically
    Estimated from the real repetitions in the training set (Methods, 'Targeted modification of orientations'). These statistics define the normal distributions from which augmentation parameters are sampled; the paper does not report the values.
  • Automatic labeling thresholds for evaluation criteria = Not reported numerically; selected via random search over 10 million combinations per exercise
    Optimized to maximize the geometric mean of F1 scores on the real dataset against physiotherapist labels (Methods, 'Automatic labeling'). These thresholds are then used to label all augmented examples.
assumptions (3)
  • domain assumption The customized OpenSim musculoskeletal model (based on Rajagopal et al.) and its inverse kinematics tool enforce biomechanically plausible joint constraints for the modified trajectories.
    Used in 'Inverse kinematics'; the paper states the model is 'customized' but does not specify the modifications, and the plausibility of the output is assumed to transfer to the augmented IMU data.
  • ad hoc to paper The Euler angle offset/scaling transformation preserves the class-relevant movement characteristics or shifts them in a way that is consistent with real variations.
    The augmentation equation (es,aug = (es[t] - es[0]) ⊙ α + β) is a per-axis affine mapping; the paper provides no evidence that this covers the natural variation space of each class beyond the t-SNE visualization.
  • ad hoc to paper The automatic labeling ruleset derived from FMS/FDE criteria, with thresholds optimized on real data, yields correct labels for augmented examples.
    The ruleset thresholds are fit to the real data and there is no external validation on independently labeled augmented examples; this is the load-bearing premise for the non-label-preserving augmentation.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Boosting Automatic Exercise Evaluation Through Musculoskeletal Simulation-Based IMU Data Augmentation." pith.science (2026). https://pith.science/paper/LOSOULAH

@misc{pith2026250524415,
  author       = {Pith},
  title        = {Pith review of: Boosting Automatic Exercise Evaluation Through Musculoskeletal Simulation-Based IMU Data Augmentation},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/LOSOULAH}},
  note         = {Machine review of arXiv:2505.24415}
}
read the original abstract

Automated evaluation of movement quality can enhance physiotherapeutic treatment and sports training by providing objective, real-time feedback. However, deep learning models that assess movements captured by inertial measurement units (IMUs) are often limited by data scarcity, class imbalance, and label ambiguity. We present a data augmentation method that generates IMU data using musculoskeletal simulations integrated with systematic modifications of movement trajectories. The approach enforces anatomically plausible kinematic constraints and enables automatic labeling by combining inverse kinematic parameters with a knowledge-based evaluation strategy. Across four datasets of varying complexity, augmented variants closely resemble real-world data and contribute to gains in classification accuracy, generalization to unseen subjects, and patient-specific fine-tuning from few examples. The magnitude of these gains varies with dataset properties, in particular class balance and label ambiguity. These findings indicate that musculoskeletal simulation-based augmentation can address common challenges faced by deep learning applications in physiotherapeutic exercise evaluation.

Figures

Figures reproduced from arXiv: 2505.24415 by the authors.

Figure 1
Figure 1. Class distribution across subjects for the, Foot Drop exercise (FDE), Deep Squat and Hurdle Step exercises. Panel (a) represents the FDE, illustrating a favorable, balanced class distribution among correct execution (CE), foot drops (FD), and weight shift supporting leg (WSSL). Panel (b) focuses on the DS, revealing a more skewed distribution overall. Most participants contributed repetitions predominantly belonging… view at source ↗
Figure 2
Figure 2. Ambiguous label assignments for the Deep Squat (DS) and Hurdle Step (HS) exercises. Illustrated are the fractions of exercise repetitions that were deemed ambiguous (i.e., repetitions with differing ratings among independent physiotherapists) for individual participants. Panel (a) shows data for the DS exercise, while panel (b) corresponds to the HS exercise. Notably, the DS already displays a moderate extent of dis… view at source ↗
Figure 3
Figure 3. Exemplary illustration of the targeted modification of orientations using a Deep Squat (DS) repetition and the corresponding Euler angle trajectory of the right femur. (a) The original Euler angle trajectories are shown as solid lines, while the augmented trajectories are represented by dashed lines. The data was scaled and adjusted with an offset, where the augmentation parameters were sampled from the correspondin… view at source ↗
Figures from the paper (5 more)
Figure 4
Figure 4. Figure 4: Labeling process for the Deep Squat (DS) as defined by Cook et al. Each exercise is characterized by multiple criteria (e.g., ‘Femur is below horizontal’ or ‘Heels are elevated’). Their fulfillment or violation determines the final exercise quality label. During develo…
Figure 5
Figure 5. Figure 5: exemplarily presents the results of dimensionality reduction using t-SNE for both FDE and HS samples. The results for the DS samples are not visualized but resemble the observations for the HS samples. In the FDE dataset, augmented examples closely cluster around the c…
Figure 6
Figure 6. Figure 6: Comparison of subject-wise performance (LOSO) on real data between a classifier trained on exclusively real data (TRTR-LOSO) and a classifier trained on a combination of real and augmented data (TRATR-LOSO). This evaluation is performed for three datasets: Foot Drop ex…
Figure 7
Figure 7. Figure 7: Comparison of subject-wise patient-specific finetuning performance with and without augmented data. The comparison takes place between a baseline trained without augmented data (TRTR-LOSO) which is finetuned using only real data (TRTR-FT) and a baseline trained on real…
Figure 8
Figure 8. Figure 8: Comparison of subject-wise performance between the baseline and the finetuned model. The comparison takes place between a baseline trained on real and augmented data (TRATR-LOSO) which is finetuned using real and augmented data (TRATR-FT). This evaluation is performed …

Discussion (0). Sign in to comment.

Forward citations

Cited by 1 Pith paper

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score. Full citation record

  1. Representing and Detecting Label Ambiguity in IMU-Based Exercise Evaluation

    cs.LG 2026-07 conditional novelty 6.0 of 10

    Auto-generated label distributions let IMU exercise networks match one-hot classification while more reliably detecting ambiguous borderline repetitions and their competing classes.

Reference graph

Works this paper leans on

33 extracted references · 22 canonical work pages · cited by 1 Pith paper

  1. [1]

    A., Hussain, M

    Ashari, A., Hamid, T. A., Hussain, M. R. & Hill, K. D. Effectiveness of Individualized Home-Based Exercise on Turning and Balance Performance Among Adults Older than 50 yrs. Am. J. Phys. Medicine & Rehabil. 95, 355–365, DOI: 10.1097/phm.0000000000000388 (2016)

    doi:10.1097/phm.0000000000000388 2016

  2. [2]

    Latham, N. K. et al. Effect of a Home-Based Exercise Program on Functional Recovery Following Rehabilitation After Hip Fracture: A Randomized Clinical Trial. JAMA 311, 700–708, DOI: 10.1001/jama.2014.469 (2014)

    doi:10.1001/jama.2014.469 2014

  3. [3]

    Y ., Janakiraman, B., Gebremeskel, B

    Gelaw, A. Y ., Janakiraman, B., Gebremeskel, B. F. & Ravichandran, H. Effectiveness of Home-based rehabilitation in improving physical function of persons with Stroke and other physical disability: A systematic review of randomized controlled trials. J. Stroke Cerebrovasc. Dis. The Off. J. Natl. Stroke Assoc.29, 104800, DOI: 10.1016/j.jstrokecerebrovasdis...

    work page arXiv 2020

  4. [4]

    E., Dennis, S., Canning, C

    Flynn, A., Allen, N. E., Dennis, S., Canning, C. G. & Preston, E. Home-based prescribed exercise improves balance-related activities in people with Parkinson’s disease and has benefits similar to centre-based exercise: a systematic review. J. Physiother. 65, 189–199, DOI: 10.1016/j.jphys.2019.08.003 (2019)

    work page doi:10.1016/j.jphys.2019.08.003 2019

  5. [5]

    & Caulfield, B

    Argent, R., Daly, A. & Caulfield, B. Patient Involvement With Home-Based Exercise Programs: Can Connected Health Interventions Influence Adherence? JMIR mHealth uHealth 6, e47, DOI: 10.2196/mhealth.8518 (2018)

    doi:10.2196/mhealth.8518 2018

  6. [6]

    Faber, M. et al. The majority are not performing home-exercises correctly two weeks after their initial instruction—an assessor-blinded study. PeerJ 3, e1102, DOI: 10.7717/peerj.1102 (2015)

    doi:10.7717/peerj.1102 2015

  7. [7]

    & Hamilton, D

    Lang, S., McLelland, C., MacDonald, D. & Hamilton, D. F. Do digital interventions increase adherence to home exercise rehabilitation? A systematic review of randomised controlled trials. Arch. Physiother. 12, 24, DOI: 10.1186/ s40945-022-00148-z (2022)

    work page 2022

  8. [8]

    & Munz, M

    Spilz, A. & Munz, M. Automatic Assessment of Functional Movement Screening Exercises with Deep Learning Architectures. Sensors 23, 5, DOI: 10.3390/s23010005 (2023). Number: 1 Publisher: Multidisciplinary Digital Publishing Institute

    doi:10.3390/s23010005 2023

Show all 33 references
  1. [9]

    Cook, G., Burton, L., Hoogenboom, B. J. & V oight, M. Functional movement screening: the use of fundamental movements as an assessment of function - part 1. Int. journal sports physical therapy 9, 396–409 (2014)

    2014

  2. [10]

    Cook, G., Burton, L., Hoogenboom, B. J. & V oight, M. Functional movement screening: the use of fundamental movements as an assessment of function-part 2. Int. journal sports physical therapy 9, 549–63 (2014)

    2014

  3. [11]

    Xing, Q.-J. et al. Functional movement screen dataset collected with two Azure Kinect depth sensors. Sci. Data 9, 104, DOI: 10.1038/s41597-022-01188-7 (2022)

    2022 doi

  4. [12]

    & Liang, J.-M

    Wu, W.-L., Lee, M.-H., Hsu, H.-T., Ho, W.-H. & Liang, J.-M. Development of an Automatic Functional Movement Screening System with Inertial Measurement Unit Sensors. Appl. Sci. 11, 96, DOI: 10.3390/app11010096 (2020). 16/18

    2020 doi

  5. [13]

    & Brown, K

    Scheurer, S., Tedesco, S., O’Flynn, B. & Brown, K. N. Comparing Person-Specific and Independent Models on Subject- Dependent and Independent Human Activity Recognition Performance. Sensors 20, 3647, DOI: 10.3390/s20133647 (2020)

    2020 doi

  6. [14]

    & Kulic, D

    Kianifar, R., Lee, A., Raina, S. & Kulic, D. Automated Assessment of Dynamic Knee Valgus and Risk of Knee Injury During the Single Leg Squat. IEEE journal translational engineering health medicine 5, 2100213, DOI: 10.1109/JTEHM. 2017.2736559 (2017)

    2017

  7. [15]

    M., Myers, C

    Sharifi Renani, M., Eustace, A. M., Myers, C. A. & Clary, C. W. The Use of Synthetic IMU Signals in the Training of Deep Learning Models Significantly Improves the Accuracy of Joint Kinematic Predictions. Sensors 21, 5876, DOI: 10.3390/s21175876 (2021)

    2021 doi

  8. [16]

    Mundt, M. et al. Estimation of Gait Mechanics Based on Simulated and Measured IMU Data Using an Artificial Neural Network. Front. Bioeng. Biotechnol. 8, DOI: 10.3389/fbioe.2020.00041 (2020)

    2020

  9. [17]

    & Amft, O

    Uhlenberg, L., Ole Haeusler, L. & Amft, O. SynHAR: Augmenting Human Activity Recognition With Synthetic Inertial Sensor Data Generated From Human Surface Models. IEEE Access 12, 194839–194858, DOI: 10.1109/ACCESS.2024. 3513477 (2024)

    2024 doi

  10. [18]

    Kwon, H. et al. IMUTube: Automatic Extraction of Virtual on-body Accelerometry from Video for Human Activity Recognition. Proc. ACM Interact. Mob. Wearable Ubiquitous Technol. 4, 87:1–87:29, DOI: 10.1145/3411841 (2020)

    2020 doi

  11. [19]

    Zolfaghari, P. et al. Sensor Data Augmentation from Skeleton Pose Sequences for Improving Human Activity Recognition. In 2024 International Conference on Activity and Behavior Computing (ABC) , 1–8, DOI: 10.1109/ABC61795.2024. 10652200 (2024)

    2024

  12. [20]

    & Gebremedhin, A

    Norgaard, S., Saeedi, R., Sasani, K. & Gebremedhin, A. H. Synthetic Sensor Data Generation for Health Applications: A Supervised Deep Learning Approach. In 2018 40th Annual International Conference of the IEEE Engineering in Medicine and Biology Society (EMBC) , 1164–1167, DOI...

    2018

  13. [21]

    & Behzadipour, S

    Mohammadzadeh, M., Ghadami, A., Taheri, A. & Behzadipour, S. cGAN-based high dimensional IMU sensor data generation for enhanced human activity recognition in therapeutic activities. Biomed. Signal Process. Control. 103, 107476, DOI: 10.1016/j.bspc.2024.107476 (2025)

    2025

  14. [22]

    & Yin, Q

    Zhao, J., Obonyo, E. & Yin, Q. Improving posture recognition among construction workers through data augmentation with generative adversarial network. IOP Conf. Series: Earth Environ. Sci. 1101, 092005, DOI: 10.1088/1755-1315/1101/ 9/092005 (2022)

    2022 doi

  15. [23]

    & Ploetz, T

    Leng, Z., Kwon, H. & Ploetz, T. Generating Virtual On-body Accelerometer Data from Virtual Textual Descriptions for Human Activity Recognition. In Proceedings of the 2023 ACM International Symposium on Wearable Computers , ISWC ’23, 39–43, DOI: 10.1145/3594738.3611361 (Associa...

    2023

  16. [24]

    Leng, Z. et al. IMUGPT 2.0: Language-Based Cross Modality Transfer for Sensor-Based Human Activity Recognition. Proc. ACM on Interactive, Mobile, Wearable Ubiquitous Technol. 8, 1–32, DOI: 10.1145/3678545 (2024)

    2024 doi

  17. [25]

    & Makris, D

    Chandrasekaran, M., Francik, J. & Makris, D. Gait Data Augmentation using Physics-Based Biomechanical Simulation, DOI: 10.48550/arXiv.2307.08092 (2023). ArXiv:2307.08092 [cs]

    2023 doi arXiv

  18. [26]

    Dorschky, E. et al. CNN-Based Estimation of Sagittal Plane Walking and Running Biomechanics From Measured and Simulated Inertial Sensor Data. Front. Bioeng. Biotechnol. 8, DOI: 10.3389/fbioe.2020.00604 (2020)

    2020

  19. [27]

    Orthokidataset

    Oppel & Spilz. Orthokidataset. tbd (2025)

    2025

  20. [28]

    C., Matheson, G

    Shultz, R., Anderson, S. C., Matheson, G. O., Marcello, B. & Besier, T. Test-Retest and Interrater Reliability of the Functional Movement Screen. J. Athl. Train. 48, 331–336, DOI: 10.4085/1062-6050-48.2.11 (2013)

    2013 doi

  21. [29]

    Madgwick, S. O. H., Harrison, A. J. L. & Vaidyanathan, R. Estimation of IMU and MARG orientation using a gradient descent algorithm. In 2011 IEEE International Conference on Rehabilitation Robotics , 1–7, DOI: 10.1109/ICORR.2011. 5975346 (2011)

    2011 doi

  22. [30]

    & Bellusci, G

    Paulich, M., Schepers, M., Rudigkeit, N. & Bellusci, G. Xsens MTw Awinda: Miniature Wireless Inertial-Magnetic Motion Tracker for Highly Accurate 3D Kinematic Applications. Tech. Rep., XSENS Technologies B.V . (2018)

    2018

  23. [31]

    Delp, S. L. et al. OpenSim: open-source software to create and analyze dynamic simulations of movement. IEEE transactions on bio-medical engineering 54, 1940–1950, DOI: 10.1109/TBME.2007.901024 (2007)

    2007

  24. [32]

    Rajagopal, A. et al. Full-Body Musculoskeletal Model for Muscle-Driven Simulation of Human Gait. IEEE Transactions on Biomed. Eng. 63, 2068–2079, DOI: 10.1109/tbme.2016.2586891 (2016). 17/18

    2016

  25. [33]

    Animating rotation with quaternion curves

    Shoemake, K. Animating rotation with quaternion curves. SIGGRAPH Comput. Graph. 19, 245–254, DOI: 10.1145/ 325165.325242 (1985). Author contributions statement A.A. contributed to conceptualization, methodology, data analysis, AI model development, carried out the experiments,...

    1985

Pith tools

Reviewed August 7, 2026 · model on record in the stance chip above.