pith. sign in

arxiv: 2606.27918 · v1 · pith:Y54KALTGnew · submitted 2026-06-26 · 💻 cs.CV · cs.AI

Every Step of the Way: Video-based Parkinsonian Turning Step Counting

Qiushuo Cheng , Jingjing Liu , Catherine Morgan , Alan Whone , Majid Mirmehdi This is my paper

Pith reviewed 2026-06-29 04:08 UTC · model grok-4.3

classification 💻 cs.CV cs.AI
keywords Parkinson's diseasestep countingvideo analysisturning impairment3D mesh recoveryoptical flowmultiple instance learninggait analysis
0
0 comments X

The pith

A passive video framework counts turning steps in Parkinson's disease more accurately than existing wearable methods.

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

The paper proposes a video-based method to count steps during turns for people with Parkinson's disease without requiring wearable devices. It derives an initial step estimate from foot movement signals obtained via 3D human mesh recovery, then refines the count by learning complementary dynamics from mesh and optical flow representations through cross attention. Videos of varying lengths are divided into clips whose motion embeddings are aggregated with multiple instance learning to predict a step count residual. Experiments on real-world PD turning datasets show consistent outperformance over prior step counting approaches. This approach would matter if true because it supports passive, device-free monitoring of motor dysfunction during daily activities.

Core claim

The authors present a coarse-to-fine video-based framework that first derives foot movement signals from 3D human mesh recovery for an initial step count estimate, then employs a motion encoder with cross attention to integrate fine-grained gait dynamics from mesh and optical flow representations, and uses multiple instance learning on partitioned video clips to predict step count residuals, achieving superior performance on real-world Parkinson's disease turning datasets.

What carries the argument

The coarse-to-fine pipeline that queries pixel-level motion cues from optical flow and mesh via cross attention using coarse foot movement signals from 3D mesh recovery, combined with multiple instance learning for residual prediction.

If this is right

  • The framework provides a convenient alternative to wearable devices for continuous daily monitoring.
  • Cross attention captures subtle parkinsonian gait dynamics that standard methods miss.
  • Multiple instance learning handles videos of arbitrary lengths by aggregating clip-wise embeddings.
  • Outperformance on real PD turning datasets indicates robustness to variable turning movements.

Where Pith is reading between the lines

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

  • The system could be deployed on home cameras or smartphones for remote symptom tracking.
  • The same motion representations might extend to quantifying other gait features such as stride length.
  • Performance on more severe shuffling cases would test the limits of the mesh recovery stage.
  • Integration with turning angle and duration measurements could yield a complete turning impairment score.

Load-bearing premise

The 3D human mesh recovery produces reliable foot movement signals even when patients exhibit atypical shuffling gait patterns.

What would settle it

A test video sequence of a Parkinson's patient turning where the method's step count deviates significantly from manual ground truth annotation while wearable sensors match the annotation.

Figures

Figures reproduced from arXiv: 2606.27918 by Alan Whone, Catherine Morgan, Jingjing Liu, Majid Mirmehdi, Qiushuo Cheng.

Figure 1
Figure 1. Figure 1: Turning steps in Parkinson’s Disease (PD). 3D mesh captures coarse-grained mo [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Overall framework for step detection. A foot movement signal [PITH_FULL_IMAGE:figures/full_fig_p005_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Cross attention in motion encoder. The horizontal axis corresponds to mesh [PITH_FULL_IMAGE:figures/full_fig_p011_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Step count estimation results on PD-FOG. (Left) Frames from an example from [PITH_FULL_IMAGE:figures/full_fig_p011_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Illustration of the correlations between the step-level turning parameters and the [PITH_FULL_IMAGE:figures/full_fig_p012_5.png] view at source ↗
read the original abstract

As a prominent symptom of Parkinson's disease (PD), turning impairment is evaluated through parameters such as turning angle, duration, and particularly, the number of steps required to complete a turn, which directly reflects motor dysfunction. Accurate step counting is challenging due to variability in real-world turning movements and atypical shuffling patterns in parkinsonian gait. Existing methods are predominantly wearable-based, requiring users to wear and manage dedicated devices, which can be inconvenient for continuous daily use. To address this, we propose a passive, video-based framework that estimates step count in a coarse-to-fine manner using diverse motion representations. Specifically, an initial step count is estimated from foot movement signals derived from 3D human mesh recovery, providing high-level motion structures. To incorporate fine-grained motion details, a motion encoder learns complementary gait dynamics from mesh and optical flow to refine the initial estimate. In this process, coarse foot movement signals query the pixel-level motion cues via cross attention to capture subtle parkinsonian gait dynamics. To handle varying video lengths, we partition each video into clips and integrate clip-wise motion embeddings via multiple instance learning (MIL) for step count residual prediction. Extensive experiments show our method consistently outperforms existing step counting methods on real-world PD turning datasets.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

2 major / 0 minor

Summary. The manuscript proposes a passive video-based method for counting steps during turning in Parkinson's disease (PD) patients. It employs a coarse-to-fine pipeline: an initial step count is derived from foot movement signals obtained via 3D human mesh recovery, which is then refined using a motion encoder that integrates mesh and optical flow features through cross-attention. Variable-length videos are handled by partitioning into clips and using multiple instance learning (MIL) to predict the step count residual. The paper claims that extensive experiments demonstrate consistent outperformance over existing step counting methods on real-world PD turning datasets.

Significance. Should the empirical claims hold after addressing validation gaps, the work could provide a practical non-contact alternative to wearables for assessing turning impairment, a clinically relevant PD motor symptom. The structured use of 3D mesh for coarse structure, cross-attention for fine gait dynamics, and MIL for variable lengths is a coherent technical response to the problem of atypical shuffling gait.

major comments (2)
  1. [Abstract / coarse-to-fine pipeline description] The central claim of consistent outperformance rests on the coarse stage producing usable initial step counts from 3D human mesh recovery even for parkinsonian shuffling (minimal lift, atypical posture). The abstract and method description supply no quantitative mesh accuracy metrics on PD data, no PD-specific fine-tuning, and no ablation isolating the coarse signal's contribution versus the cross-attention refinement. If the initial signals are noisy or biased, the MIL residual prediction cannot guarantee the headline performance.
  2. [Abstract] The abstract asserts 'extensive experiments' and 'consistent outperformance' yet supplies no dataset sizes, baseline details, metrics, error bars, or exclusion criteria. Without these the central empirical claim cannot be evaluated from the provided text.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive comments, which highlight important aspects of validation and clarity. We address each major comment below and will revise the manuscript to strengthen the presentation of our results.

read point-by-point responses

  1. Referee: [Abstract / coarse-to-fine pipeline description] The central claim of consistent outperformance rests on the coarse stage producing usable initial step counts from 3D human mesh recovery even for parkinsonian shuffling (minimal lift, atypical posture). The abstract and method description supply no quantitative mesh accuracy metrics on PD data, no PD-specific fine-tuning, and no ablation isolating the coarse signal's contribution versus the cross-attention refinement. If the initial signals are noisy or biased, the MIL residual prediction cannot guarantee the headline performance.

    Authors: We agree that explicit validation of the coarse stage is valuable for readers. The full manuscript demonstrates the end-to-end superiority on real-world PD turning videos, but does not report per-frame mesh accuracy numbers on the PD cohort or an ablation that isolates the coarse foot-signal contribution. We will add both: (i) quantitative mesh reconstruction metrics computed on our PD dataset and (ii) an ablation that removes the coarse stage (or replaces it with a constant) while keeping the cross-attention and MIL components fixed. The off-the-shelf mesh model is retained without PD-specific fine-tuning because it already supplies usable high-level foot trajectories for the subsequent refinement; the performance gap versus purely image-based or wearable baselines supports this design choice. revision: yes

  2. Referee: [Abstract] The abstract asserts 'extensive experiments' and 'consistent outperformance' yet supplies no dataset sizes, baseline details, metrics, error bars, or exclusion criteria. Without these the central empirical claim cannot be evaluated from the provided text.

    Authors: The current abstract is intentionally concise. We will expand it in the revision to report the number of videos and subjects, the main competing methods, the primary metrics (with standard deviations), and a brief statement on exclusion criteria, thereby allowing readers to assess the empirical claims directly from the abstract. revision: yes

Circularity Check

0 steps flagged

No circularity: standard supervised pipeline with empirical validation

full rationale

The paper describes a coarse-to-fine video pipeline that extracts initial step counts via external 3D mesh recovery, refines via cross-attention motion encoding, and predicts residuals with MIL. No equations, self-definitions, or fitted parameters are shown to reduce the reported outperformance metrics to quantities defined by the method's own inputs. Claims rest on experiments against external PD turning datasets rather than any derivation that collapses by construction. This is a conventional supervised learning setup with no load-bearing self-citation chains or ansatz smuggling.

Axiom & Free-Parameter Ledger

2 free parameters · 2 axioms · 0 invented entities

Abstract-only review limits visibility; the pipeline implicitly assumes standard 3D mesh recovery works on PD gait and that clip partitioning plus MIL can aggregate variable-length inputs without loss of count accuracy.

free parameters (2)
  • clip partitioning strategy
    Chosen to handle varying video lengths; exact clip duration and overlap not stated.
  • cross-attention hyperparameters
    Learned weights and embedding dimensions fitted during training.
axioms (2)
  • domain assumption 3D human mesh recovery produces usable foot trajectories on parkinsonian shuffling gait
    Invoked by the initial step count estimation step.
  • domain assumption Optical flow supplies complementary fine-grained cues not captured by mesh alone
    Basis for the refinement stage.

pith-pipeline@v0.9.1-grok · 5759 in / 1418 out tokens · 45631 ms · 2026-06-29T04:08:49.604971+00:00 · methodology

discussion (0)

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

Reference graph

Works this paper leans on

61 extracted references · 1 canonical work pages

  1. [1]

    A systematic review of wearable patient monitoring systems–current challenges and opportunities for clinical adoption.Journal of medical systems, 41(7): 115, 2017

    Mirza Mansoor Baig, Hamid GholamHosseini, Aasia A Moqeem, Farhaan Mirza, and Maria Lindén. A systematic review of wearable patient monitoring systems–current challenges and opportunities for clinical adoption.Journal of medical systems, 41(7): 115, 2017

    2017

  2. [2]

    Step counting: a review of measurement considerations and health-related applications

    David R Bassett Jr, Lindsay P Toth, Samuel R LaMunion, and Scott E Crouter. Step counting: a review of measurement considerations and health-related applications. Sports Medicine, 47(7):1303–1315, 2017

    2017

  3. [3]

    Precise step counting algorithm for pedestrians us- ing ultra-low-cost foot-mounted accelerometer.Engineering Applications of Artificial Intelligence, 150:110619, 2025

    Jingxue Bi, Jianhui Wang, Baoguo Yu, Guobiao Yao, Yunjia Wang, Hongji Cao, Lu Huang, and Huaqiao Xing. Precise step counting algorithm for pedestrians us- ing ultra-low-cost foot-mounted accelerometer.Engineering Applications of Artificial Intelligence, 150:110619, 2025

    2025

  4. [4]

    Prospective assessment of falls in parkinson’s disease.Journal of neurology, 248(11):950–958, 2001

    Bastiaan R Bloem, Yvette AM Grimbergen, Monique Cramer, Mirjam Willemsen, and Aeilko H Zwinderman. Prospective assessment of falls in parkinson’s disease.Journal of neurology, 248(11):950–958, 2001

    2001

  5. [5]

    Parkinson’s disease.The Lancet, 397(10291):2284–2303, 2021

    Bastiaan R Bloem, Michael S Okun, and Christine Klein. Parkinson’s disease.The Lancet, 397(10291):2284–2303, 2021

    2021

  6. [6]

    Carlotta Caramia, Diego Torricelli, Maurizio Schmid, Adriana Muñoz-Gonzalez, Jose Gonzalez-Vargas, Francisco Grandas, and Jose L Pons. Imu-based classification of parkinson’s disease from gait: A sensitivity analysis on sensor location and feature selection.IEEE journal of biomedical and health informatics, 22(6):1765–1774, 2018

    2018

  7. [7]

    Your turn: At home turning angle estimation for parkin- son’s disease severity assessment.Artificial Intelligence in Medicine, 167:103194, 2025

    Qiushuo Cheng, Catherine Morgan, Arindam Sikdar, Alessandro Masullo, Alan Whone, and Majid Mirmehdi. Your turn: At home turning angle estimation for parkin- son’s disease severity assessment.Artificial Intelligence in Medicine, 167:103194, 2025

    2025

  8. [8]

    Re- liability and validity of current computer vision based motion capture systems in gait analysis: A systematic review.Gait & Posture, 120:150–160, 2025

    Xingye Cheng, Yiran Jiao, Rebecca M Meiring, Bo Sheng, and Yanxin Zhang. Re- liability and validity of current computer vision based motion capture systems in gait analysis: A systematic review.Gait & Posture, 120:150–160, 2025. 14CHENG ET AL: VIDEO-BASED PARKINSONIAN TURNING STEP COUNTING

    2025

  9. [9]

    Pose-based gait analysis for diagnosis of parkinson’s disease.Algorithms, 15(12):474, 2022

    Tee Connie, Timilehin B Aderinola, Thian Song Ong, Michael Kah Ong Goh, Bayu Erfianto, and Bedy Purnama. Pose-based gait analysis for diagnosis of parkinson’s disease.Algorithms, 15(12):474, 2022

    2022

  10. [10]

    Falls, 1994

    Steven R Cummings and Michael C Nevitt. Falls, 1994

    1994

  11. [11]

    Daily steps and health outcomes in adults: a systematic review and dose-response meta-analysis.The Lancet Public Health, 10(8):e668–e681, 2025

    Ding Ding, Binh Nguyen, Tracy Nau, Mengyun Luo, Borja del Pozo Cruz, Paddy C Dempsey, Zachary Munn, Barbara J Jefferis, Cathie Sherrington, Elizabeth A Calleja, et al. Daily steps and health outcomes in adults: a systematic review and dose-response meta-analysis.The Lancet Public Health, 10(8):e668–e681, 2025

    2025

  12. [12]

    Counting out time: Class agnostic video repetition counting in the wild

    Debidatta Dwibedi, Yusuf Aytar, Jonathan Tompson, Pierre Sermanet, and Andrew Zisserman. Counting out time: Class agnostic video repetition counting in the wild. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 10387–10396, 2020

    2020

  13. [13]

    Continuous monitoring of turning in patients with movement disability.Sensors, 14(1):356–369, 2013

    Mahmoud El-Gohary, Sean Pearson, James McNames, Martina Mancini, Fay Horak, Sabato Mellone, and Lorenzo Chiari. Continuous monitoring of turning in patients with movement disability.Sensors, 14(1):356–369, 2013

    2013

  14. [14]

    “on” state freezing of gait in parkinson disease: a paradoxical levodopa-induced complica- tion.Neurology, 78(7):454–457, 2012

    AJ Espay, A Fasano, BFL Van Nuenen, MM Payne, AH Snijders, and BR Bloem. “on” state freezing of gait in parkinson disease: a paradoxical levodopa-induced complica- tion.Neurology, 78(7):454–457, 2012

    2012

  15. [15]

    Deep neural networks for human activity recognition with wearable sensors: Leave-one-subject-out cross-validation for model selection.Ieee Access, 8:133982–133994, 2020

    Davoud Gholamiangonabadi, Nikita Kiselov, and Katarina Grolinger. Deep neural networks for human activity recognition with wearable sensors: Leave-one-subject-out cross-validation for model selection.Ieee Access, 8:133982–133994, 2020

    2020

  16. [16]

    Christopher G Goetz, Barbara C Tilley, Stephanie R Shaftman, Glenn T Stebbins, Stan- ley Fahn, Pablo Martinez-Martin, Werner Poewe, Cristina Sampaio, Matthew B Stern, Richard Dodel, et al. Movement disorder society-sponsored revision of the unified parkinson’s disease rating scale (mds-updrs): scale presentation and clinimetric testing results.Movement di...

    2008

  17. [17]

    Momentum con- trast for unsupervised visual representation learning

    Kaiming He, Haoqi Fan, Yuxin Wu, Saining Xie, and Ross Girshick. Momentum con- trast for unsupervised visual representation learning. InProceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 9729–9738, 2020

    2020

  18. [18]

    Gait and balance in parkinson’s disease subtypes: objective measures and classification con- siderations.Journal of neurology, 261(12):2401–2410, 2014

    Talia Herman, Aner Weiss, Marina Brozgol, Nir Giladi, and Jeffrey M Hausdorff. Gait and balance in parkinson’s disease subtypes: objective measures and classification con- siderations.Journal of neurology, 261(12):2401–2410, 2014

    2014

  19. [19]

    Automated gait analysis based on a marker-free pose estimation model.Sensors, 23(14):6489, 2023

    Chang Soon Tony Hii, Kok Beng Gan, Nasharuddin Zainal, Norlinah Mo- hamed Ibrahim, Shahrul Azmin, Siti Hajar Mat Desa, Bart van de Warrenburg, and Huay Woon You. Automated gait analysis based on a marker-free pose estimation model.Sensors, 23(14):6489, 2023

    2023

  20. [20]

    Parkinsonism: onset, progression, and mortal- ity.Neurology, 17(5):427–427, 1967

    Margaret M Hoehn and Melvin D Yahr. Parkinsonism: onset, progression, and mortal- ity.Neurology, 17(5):427–427, 1967. CHENG ET AL: VIDEO-BASED PARKINSONIAN TURNING STEP COUNTING15

    1967

  21. [21]

    Pedestrian dead reckoning with wearable sensors: A systematic review.IEEE Sensors Journal, 21(1):143–152, 2020

    Xinyu Hou and Jeroen Bergmann. Pedestrian dead reckoning with wearable sensors: A systematic review.IEEE Sensors Journal, 21(1):143–152, 2020

    2020

  22. [22]

    A narrative review of turning deficits in people with parkinson’s disease.Disability and rehabilitation, 37 (15):1382–1389, 2015

    Sophia Hulbert, Ann Ashburn, Lisa Robert, and Geert Verheyden. A narrative review of turning deficits in people with parkinson’s disease.Disability and rehabilitation, 37 (15):1382–1389, 2015

    2015

  23. [23]

    Attention-based deep multiple instance learning

    Maximilian Ilse, Jakub Tomczak, and Max Welling. Attention-based deep multiple instance learning. InInternational conference on machine learning, pages 2127–2136. PMLR, 2018

    2018

  24. [24]

    Tsm: Temporal shift module for efficient video understanding

    Ji Lin, Chuang Gan, and Song Han. Tsm: Temporal shift module for efficient video understanding. InProceedings of the IEEE/CVF international conference on computer vision, pages 7083–7093, 2019

    2019

  25. [25]

    A novel robust step detection algorithm for foot-mounted imu.IEEE Sensors Journal, 21(4):5331–5339, 2020

    Xin Liu, Ning Li, Geng Xu, and Yonggang Zhang. A novel robust step detection algorithm for foot-mounted imu.IEEE Sensors Journal, 21(4):5331–5339, 2020

    2020

  26. [26]

    Smpl: A skinned multi-person linear model

    Matthew Loper, Naureen Mahmood, Javier Romero, Gerard Pons-Moll, and Michael J Black. Smpl: A skinned multi-person linear model. InSeminal Graphics Papers: Pushing the Boundaries, Volume 2, pages 851–866. 2023

    2023

  27. [27]

    Multi-activity step counting al- gorithm using deep learning foot flat detection with an imu inside the sole of a shoe

    Quentin Lucot, Erwan Beurienne, and Michel Behr. Multi-activity step counting al- gorithm using deep learning foot flat detection with an imu inside the sole of a shoe. Sensors, 24(21):6927, 2024

    2024

  28. [28]

    On-shoe wearable sensors for gait and turning assessment of patients with parkinson’s disease.IEEE transactions on biomedical engineering, 60(1):155–158, 2012

    Benoit Mariani, Mayté Castro Jiménez, François JG Vingerhoets, and Kamiar Aminian. On-shoe wearable sensors for gait and turning assessment of patients with parkinson’s disease.IEEE transactions on biomedical engineering, 60(1):155–158, 2012

    2012

  29. [29]

    A framework for multiple-instance learning

    Oded Maron and Tomás Lozano-Pérez. A framework for multiple-instance learning. Advances in neural information processing systems, 10, 1997

    1997

  30. [30]

    Validation of an algorithm for detecting turning in people with cognitive impairment, considering dementia disease subtype.Gait & posture, 118:141–147, 2025

    Ríona Mc Ardle, Leigh J Ryan, Rana Zia Ur Rehman, Emily Dignan, Abbie Thomp- son, Silvia Del Din, Brook Galna, Alan J Thomas, Lynn Rochester, and Lisa Alcock. Validation of an algorithm for detecting turning in people with cognitive impairment, considering dementia disease subtype.Gait & posture, 118:141–147, 2025

    2025

  31. [31]

    Gait impairments in parkinson’s disease.The Lancet Neurology, 18(7):697–708, 2019

    Anat Mirelman, Paolo Bonato, Richard Camicioli, Terry D Ellis, Nir Giladi, Jamie L Hamilton, Chris J Hass, Jeffrey M Hausdorff, Elisa Pelosin, and Quincy J Almeida. Gait impairments in parkinson’s disease.The Lancet Neurology, 18(7):697–708, 2019

    2019

  32. [32]

    Mobile device accuracy for step counting across age groups.JMIR mHealth and uHealth, 5(6):e7870, 2017

    François Modave, Yi Guo, Jiang Bian, Matthew J Gurka, Alice Parish, Megan D Smith, Alexandra M Lee, and Thomas W Buford. Mobile device accuracy for step counting across age groups.JMIR mHealth and uHealth, 5(6):e7870, 2017

    2017

  33. [33]

    Understanding how people with parkinson’s disease turn in gait from a real-world in-home dataset.Parkinsonism & related disorders, 105: 114–122, 2022

    Catherine Morgan, Jack Jameson, Ian Craddock, Emma L Tonkin, George Oikonomou, Hanna Kristiina Isotalus, Farnoosh Heidarivincheh, Ryan McConville, Gregory JL Tourte, Kirsi M Kinnunen, et al. Understanding how people with parkinson’s disease turn in gait from a real-world in-home dataset.Parkinsonism & related disorders, 105: 114–122, 2022. 16CHENG ET AL: ...

    2022

  34. [34]

    A multimodal dataset of real world mobility activities in parkinson’s disease.Scientific data, 10(1):918, 2023

    Catherine Morgan, Emma L Tonkin, Alessandro Masullo, Ferdian Jovan, Arindam Sik- dar, Pushpajit Khaire, Majid Mirmehdi, Ryan McConville, Gregory JL Tourte, Alan Whone, et al. A multimodal dataset of real world mobility activities in parkinson’s disease.Scientific data, 10(1):918, 2023

    2023

  35. [35]

    Step-count outcomes of 13 different activity trackers: results from labora- tory and free-living experiments.Gait & Posture, 98:24–33, 2022

    Takashi Nakagata, Haruka Murakami, Ryoko Kawakami, Julien Tripette, Satoshi Nakae, Yosuke Yamada, Kazuko Ishikawa-Takata, Shigeho Tanaka, and Motohiko Miyachi. Step-count outcomes of 13 different activity trackers: results from labora- tory and free-living experiments.Gait & Posture, 98:24–33, 2022

    2022

  36. [36]

    Multi-domain cop feature analysis of func- tional mobility for parkinson’s disease detection using wearable pressure insoles.Sen- sors, 25(18):5859, 2025

    Thathsara Nanayakkara, HMKKMB Herath, Hadi Sedigh Malekroodi, Nuwan Madu- sanka, Myunggi Yi, and Byeong-il Lee. Multi-domain cop feature analysis of func- tional mobility for parkinson’s disease detection using wearable pressure insoles.Sen- sors, 25(18):5859, 2025

    2025

  37. [37]

    Validity of instrumented insoles for step counting, posture and activity recognition: a systematic review.Sen- sors, 19(11):2438, 2019

    Armelle M Ngueleu, Andréanne K Blanchette, Désirée Maltais, Hélène Moffet, Brad- ford J McFadyen, Laurent Bouyer, and Charles S Batcho. Validity of instrumented insoles for step counting, posture and activity recognition: a systematic review.Sen- sors, 19(11):2438, 2019

    2019

  38. [38]

    Minh H Pham, Morad Elshehabi, Linda Haertner, Silvia Del Din, Karin Srulijes, Tanja Heger, Matthis Synofzik, Markus A Hobert, Gert S Faber, Clint Hansen, et al. Valida- tion of a step detection algorithm during straight walking and turning in patients with parkinson’s disease and older adults using an inertial measurement unit at the lower back.Frontiers ...

    2017

  39. [39]

    Minh H Pham, Morad Elshehabi, Linda Haertner, Tanja Heger, Markus A Hobert, Gert S Faber, Dina Salkovic, Joaquim J Ferreira, Daniela Berg, Álvaro Sanchez-Ferro, et al. Algorithm for turning detection and analysis validated under home-like condi- tions in patients with parkinson’s disease and older adults using a 6 degree-of-freedom inertial measurement un...

    2017

  40. [40]

    The timed “up & go”: a test of basic func- tional mobility for frail elderly persons.Journal of the American geriatrics Society, 39 (2):142–148, 1991

    Diane Podsiadlo and Sandra Richardson. The timed “up & go”: a test of basic func- tional mobility for frail elderly persons.Journal of the American geriatrics Society, 39 (2):142–148, 1991

    1991

  41. [41]

    Flowseek: optical flow made easier with depth founda- tion models and motion bases

    Matteo Poggi and Fabio Tosi. Flowseek: optical flow made easier with depth founda- tion models and motion bases. InProceedings of the IEEE/CVF International Confer- ence on Computer Vision, pages 5667–5679, 2025

    2025

  42. [42]

    Rana Zia Ur Rehman, Philipp Klocke, Sofia Hryniv, Brook Galna, Lynn Rochester, Sil- via Del Din, and Lisa Alcock. Turning detection during gait: algorithm validation and influence of sensor location and turning characteristics in the classification of parkin- son’s disease.Sensors, 20(18):5377, 2020

    2020

  43. [43]

    Humor: 3d human motion model for robust pose estimation

    Davis Rempe, Tolga Birdal, Aaron Hertzmann, Jimei Yang, Srinath Sridhar, and Leonidas J Guibas. Humor: 3d human motion model for robust pose estimation. InProceedings of the IEEE/CVF international conference on computer vision, pages 11488–11499, 2021. CHENG ET AL: VIDEO-BASED PARKINSONIAN TURNING STEP COUNTING17

    2021

  44. [44]

    Caroline Ribeiro De Souza, Runfeng Miao, Júlia Ávila De Oliveira, Andrea Cristina De Lima-Pardini, Débora Fragoso De Campos, Carla Silva-Batista, Luis Teixeira, So- laiman Shokur, Bouri Mohamed, and Daniel Boari Coelho. A public data set of videos, inertial measurement unit, and clinical scales of freezing of gait in individuals with parkinson’s disease d...

    2022

  45. [45]

    Textured foot orthotics on dynamic stability and turning performance in parkinson’s disease.Journal of Motor Behavior, 52(4):396– 403, 2020

    Kelly A Robb and Stephen D Perry. Textured foot orthotics on dynamic stability and turning performance in parkinson’s disease.Journal of Motor Behavior, 52(4):396– 403, 2020

    2020

  46. [46]

    Automation of the timed-up-and-go test using a conventional video camera.IEEE journal of biomed- ical and health informatics, 24(4):1196–1205, 2019

    Patrick Savoie, James AD Cameron, Mary E Kaye, and Erik J Scheme. Automation of the timed-up-and-go test using a conventional video camera.IEEE journal of biomed- ical and health informatics, 24(4):1196–1205, 2019

    2019

  47. [47]

    A comprehensive survey on gait analysis: History, parameters, approaches, pose estimation, and future work.Artificial Intelligence in Medicine, 129:102314, 2022

    Dimple Sethi, Sourabh Bharti, and Chandra Prakash. A comprehensive survey on gait analysis: History, parameters, approaches, pose estimation, and future work.Artificial Intelligence in Medicine, 129:102314, 2022

    2022

  48. [48]

    World-grounded human motion recovery via gravity- view coordinates

    Zehong Shen, Huaijin Pi, Yan Xia, Zhi Cen, Sida Peng, Zechen Hu, Hujun Bao, Ruizhen Hu, and Xiaowei Zhou. World-grounded human motion recovery via gravity- view coordinates. InSIGGRAPH Asia 2024 Conference Papers, pages 1–11, 2024

    2024

  49. [50]

    Quantitative gait analysis us- ing a pose-estimation algorithm with a single 2d-video of parkinson’s disease patients

    Jung Hwan Shin, Ri Yu, Jed Noel Ong, Chan Young Lee, Seung Ho Jeon, Hwanpil Park, Han-Joon Kim, Jehee Lee, and Beomseok Jeon. Quantitative gait analysis us- ing a pose-estimation algorithm with a single 2d-video of parkinson’s disease patients. Journal of Parkinson’s Disease, 11(3):1271–1283, 2021

    2021

  50. [51]

    Every shot counts: Using exemplars for repetition counting in videos

    Saptarshi Sinha, Alexandros Stergiou, and Dima Damen. Every shot counts: Using exemplars for repetition counting in videos. InProceedings of the Asian Conference on Computer Vision, pages 3056–3073, 2024

    2024

  51. [52]

    Head-pelvis cou- pling is increased during turning in patients with p arkinson’s disease and freezing of gait.Movement Disorders, 28(5):619–625, 2013

    Joke Spildooren, Sarah Vercruysse, Elke Heremans, Brook Galna, Jochen Vandenboss- che, Kaat Desloovere, Wim Vandenberghe, and Alice Nieuwboer. Head-pelvis cou- pling is increased during turning in patients with p arkinson’s disease and freezing of gait.Movement Disorders, 28(5):619–625, 2013

    2013

  52. [53]

    Two-dimensional video-based analysis of human gait using pose estimation.PLoS computational biology, 17(4): e1008935, 2021

    Jan Stenum, Cristina Rossi, and Ryan T Roemmich. Two-dimensional video-based analysis of human gait using pose estimation.PLoS computational biology, 17(4): e1008935, 2021

    2021

  53. [54]

    Clinical gait analysis using video-based pose estimation: Multiple perspectives, clinical popu- lations, and measuring change.PLOS Digital Health, 3(3):e0000467, 2024

    Jan Stenum, Melody M Hsu, Alexander Y Pantelyat, and Ryan T Roemmich. Clinical gait analysis using video-based pose estimation: Multiple perspectives, clinical popu- lations, and measuring change.PLOS Digital Health, 3(3):e0000467, 2024. 18CHENG ET AL: VIDEO-BASED PARKINSONIAN TURNING STEP COUNTING

    2024

  54. [55]

    Periodic motion detection and estimation via space-time sampling

    Ashwin Thangali and Stan Sclaroff. Periodic motion detection and estimation via space-time sampling. In2005 Seventh IEEE Workshops on Applications of Computer Vision (WACV/MOTION’05)-Volume 1, volume 2, pages 176–182. IEEE, 2005

    2005

  55. [56]

    Validity of different activity monitors to count steps in an inpa- tient rehabilitation setting.Physical therapy, 97(5):581–588, 2017

    Daniel Treacy, Leanne Hassett, Karl Schurr, Sakina Chagpar, Serene S Paul, and Catherine Sherrington. Validity of different activity monitors to count steps in an inpa- tient rehabilitation setting.Physical therapy, 97(5):581–588, 2017

    2017

  56. [57]

    Vision-based motion capture for the gait analysis of neurodegenerative diseases: A review.Gait & posture, 112:95–107, 2024

    David Sing Yee Vun, Robert Bowers, and Anthony McGarry. Vision-based motion capture for the gait analysis of neurodegenerative diseases: A review.Gait & posture, 112:95–107, 2024

    2024

  57. [58]

    Prompthmr: Promptable human mesh recovery

    Yufu Wang, Yu Sun, Priyanka Patel, Kostas Daniilidis, Michael J Black, and Muhammed Kocabas. Prompthmr: Promptable human mesh recovery. InProceedings of the computer vision and pattern recognition conference, pages 1148–1159, 2025

    2025

  58. [59]

    Poserac: Pose saliency transformer for repetitive action counting.arXiv preprint arXiv:2303.08450, 2023

    Ziyu Yao, Xuxin Cheng, and Yuexian Zou. Poserac: Pose saliency transformer for repetitive action counting.arXiv preprint arXiv:2303.08450, 2023

    work page arXiv 2023

  59. [60]

    Cris Zampieri, Arash Salarian, Patricia Carlson-Kuhta, Kamiar Aminian, John G Nutt, and Fay B Horak. The instrumented timed up and go test: potential outcome measure for disease modifying therapies in parkinson’s disease.Journal of Neurology, Neuro- surgery & Psychiatry, 81(2):171–176, 2010

    2010

  60. [61]

    Cola: Weakly- supervised temporal action localization with snippet contrastive learning

    Can Zhang, Meng Cao, Dongming Yang, Jie Chen, and Yuexian Zou. Cola: Weakly- supervised temporal action localization with snippet contrastive learning. InProceed- ings of the IEEE/CVF conference on computer vision and pattern recognition, pages 16010–16019, 2021

    2021

  61. [62]

    Reducing footskate in human motion reconstruction with ground contact constraints

    Yuliang Zou, Jimei Yang, Duygu Ceylan, Jianming Zhang, Federico Perazzi, and Jia- Bin Huang. Reducing footskate in human motion reconstruction with ground contact constraints. InProceedings of the IEEE/CVF Winter Conference on Applications of Computer Vision, pages 459–468, 2020

    2020