K2MUSE: A human lower-limb multimodal walking dataset spanning task and acquisition variability for rehabilitation robotics

Bi Zhang; Jian Huang; Jiwei Li; Juanjuan Zhang; Lianqing Liu; Wanxin Chen; Weiguang Huo; Xiaowei Tan; Xingang Zhao; Zhaoyuan Liu

arxiv: 2504.14602 · v3 · submitted 2025-04-20 · 💻 cs.RO · cs.AI· cs.HC

K2MUSE: A human lower-limb multimodal walking dataset spanning task and acquisition variability for rehabilitation robotics

Jiwei Li , Bi Zhang , Xiaowei Tan , Wanxin Chen , Zhaoyuan Liu , Juanjuan Zhang , Weiguang Huo , Jian Huang

show 2 more authors

Lianqing Liu Xingang Zhao

This is my paper

Pith reviewed 2026-05-22 19:21 UTC · model grok-4.3

classification 💻 cs.RO cs.AIcs.HC

keywords lower-limb multimodal datasetrehabilitation roboticsgait analysissurface electromyographyamplitude mode ultrasoundwalking on inclinesmuscle fatigueolder adults locomotion

0 comments

The pith

K2MUSE provides multimodal lower-limb walking data from young and older adults across speeds, inclines, and non-ideal sensor conditions.

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

This paper presents the K2MUSE dataset to supply the multimodal measurements and scale of gait samples missing from prior lower-limb collections used in rehabilitation robotics. Current resources overlook the effects of real acquisition interference such as muscle fatigue, electrode movement, and day-to-day differences that affect sensor-based control. The new collection records synchronized kinematic, kinetic, surface electromyography, and amplitude-mode ultrasound signals from forty-two participants while they walk at multiple speeds and inclines, including under fatigue and sensor-shift conditions. Structured documentation, preprocessing code, and example scripts accompany the release to support immediate use in data-driven modeling.

Core claim

The K2MUSE dataset comprises lower-limb multimodal recordings from 30 young adults and 12 older adults, including kinematic and kinetic data from a Vicon system and instrumented treadmill together with sEMG and AUS signals from thirteen bilateral muscles, captured across inclines of 0°, ±5°, and ±10°, speeds of 0.5 m/s, 1.0 m/s, and 1.5 m/s, and representative non-ideal conditions of muscle fatigue, electrode shifts, and interday differences.

What carries the argument

The K2MUSE dataset itself, a synchronized collection of kinematic, kinetic, sEMG, and amplitude-mode ultrasound measurements gathered under controlled task and acquisition variability.

If this is right

Supports training of data-driven controllers that maintain performance when sEMG electrodes shift or muscles fatigue during prolonged sessions.
Enables direct comparison of gait patterns between young and older adults on the same inclines and speeds.
Supplies benchmark recordings for testing wearable-sensor algorithms that must tolerate interday signal changes.
Provides synchronized ground-truth labels for developing fusion methods that combine ultrasound imaging with electromyography.

Where Pith is reading between the lines

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

The dataset could be used to test whether fatigue detection modules improve robot assistance timing in clinical therapy sessions.
Researchers might extend the recordings to study how older-adult data affects generalization of exoskeleton controllers trained mainly on young-adult samples.
Future work could add real-time feedback loops that use the dataset to simulate sensor drift and train corrective models.

Load-bearing premise

The collected multimodal signals stay sufficiently synchronized and representative of real clinical use cases to enable downstream improvements in rehabilitation robot performance.

What would settle it

An experiment in which models trained on K2MUSE fail to show higher accuracy or robustness in predicting intended motion or controlling a lower-limb exoskeleton compared with models trained on existing datasets when tested under fatigue or electrode-shift conditions.

Figures

Figures reproduced from arXiv: 2504.14602 by Bi Zhang, Jian Huang, Jiwei Li, Juanjuan Zhang, Lianqing Liu, Wanxin Chen, Weiguang Huo, Xiaowei Tan, Xingang Zhao, Zhaoyuan Liu.

**Figure 1.** Figure 1: Experiments were conducted in the biomechanics laboratory. (a) The experimental scene shows a participant equipped with all the devices: a motion capture system, a treadmill with embedded force plates, an sEMG system, and an AUS device. (b) Participants performed experiments on a treadmill under diverse conditions, including different ascending and descending ramps and walking speeds. In the fatigue-induce… view at source ↗

**Figure 2.** Figure 2: The modified marker set for motion capture. The markers were attached to the lower limbs in a generally symmetrical arrangement, with the markers on the left side shown. Markers marked with ‘*’ were defined according to the Plug-in Gait lower body model, which implements the Conventional Gait Model. Detailed marker placement instructions for the Plug-in Gait lower body model can be found in the Plug-in Gai… view at source ↗

**Figure 3.** Figure 3: The sEMG sensors and AUS transducers were attached to the participants’ skin. The channels of different instrumentation are highlighted in different colors for easy distinction. The symbol ‘#’ corresponds to the channel numbers of different devices. markers were placed on the lower body through manual palpation. All the markers were securely fixed to the skin using a combination of double-sided tape and PU… view at source ↗

**Figure 4.** Figure 4: Experimental setups for muscle fatigue and electrode shifts. (a) Fatigue-induced lower limb exercises, including dorsiflexion/plantar flexion and squats, with a barbell held in the hands. (b) Experimental setup simulating electrode shifts, where different electrode pairs and transducers correspond to initial positions and four shift directions. shoulder-width apart on the ground and held a barbell plate (1… view at source ↗

**Figure 5.** Figure 5: Data organization outlines (a) the folder structure of the dataset and (b) the structured ’P*.mat’ file for the participant [PITH_FULL_IMAGE:figures/full_fig_p012_5.png] view at source ↗

**Figure 6.** Figure 6: During level-ground walking at 1.0 m/s, the representative AUS and sEMG data of the left rectus femoris (RF), along with the left hip joint angle data, were recorded. The insets on the right side of the AUS data present representative raw AUS data captured at specific time frames. 5.5.1 Muscle fatigue. Muscle fatigue can lead to feature drift in human-machine interfaces, as sEMG-based features are sensitiv… view at source ↗

**Figure 7.** Figure 7: Joint angles and moments during ideal condition experiments. Two consecutive heel strikes correspond to 0% and 100% of the gait cycle. The ambulation mode is annotated at the top. The solid lines indicate the average trajectory across all participants. The shaded regions correspond to the standard deviation. 5.6 Joint angle prediction To assess the performance of decoding lower limb movements using sEMG an… view at source ↗

**Figure 8.** Figure 8: The evaluation of repeatability. The circles represent the R2 between the average angle and joint angles for each participant across different ambulation modes. The box represents the distribution of R2 for each movement. Median Frequency [Hz]Mean Frequency [Hz] [PITH_FULL_IMAGE:figures/full_fig_p016_8.png] view at source ↗

**Figure 10.** Figure 10: For the five shift positions: (a) the mean of the maximum MAV across all gait cycles for sEMG signals recorded from four muscles of the left leg; (b) MAV variation of the LG in the left leg throughout the gait cycle. The solid lines represent the average values across all cycles, whereas the shaded regions denote the standard deviation. the dimension Yang et al. (2022). We employed 5-fold crossvalidation… view at source ↗

**Figure 9.** Figure 9: Violin plots depicting the distributions of the median frequency and mean frequency of RF in the left leg during gradual muscle fatigue trials. filtering, envelope detection, and log compression Zeng et al. (2021). The frames were then segmented into a series of windows, each containing 20 sample points. The first and last 20 points were discarded prior to segmentation, as they typically do not contain val… view at source ↗

**Figure 11.** Figure 11: For ideal conditions and inter-day differences: (a) The variation in MAV values of the left leg LG throughout the gait cycle. The solid lines represent the average values across all gait cycles from five walking trials, whereas the shaded regions indicate the standard deviation. (b) Comparison of the Euclidean distance of MAV values for the left leg LG under two conditions: intra-condition trial compariso… view at source ↗

**Figure 12.** Figure 12: Regression results for angle estimation. (a), (b), and (c) show RMSE with inputs of sEMG, MSD, and SFO features, respectively. (d), (e), and (f) show RMSE with inputs of sEMG&MSD, sEMG&SFO, and MSD&SFO features, respectively. The height of the bars represents the mean value across all participants, whereas the error bars indicate the standard deviation. (g), (h), and (i) represent RMSE for electrode shift… view at source ↗

**Figure 13.** Figure 13: (e) and (f), without relying on manual tuning or predefined control laws, these controllers can be generalized to a wider range of tasks as the dataset expands, increasing their adaptability and robustness in real-world applications Molinaro et al. (2024a); Luo et al. (2024). 5.9 Data limitations Given that the dataset involved multiple acquisition systems and extensive data collection sessions, only data… view at source ↗

read the original abstract

The natural interaction and control performance of lower limb rehabilitation robots are closely linked to biomechanical information from various human locomotion activities. Multidimensional human motion data significantly deepen the understanding of the complex mechanisms governing neuromuscular alterations, thereby facilitating the development and application of rehabilitation robots in multifaceted real-world environments. However, existing lower limb datasets are inadequate for supplying the essential multimodal data and large-scale gait samples necessary for the development of effective data-driven approaches, and the significant effects of acquisition interference in real applications are neglected. To fill this gap, we present the K2MUSE dataset, which includes a comprehensive collection of multimodal data, comprising kinematic, kinetic, amplitude mode ultrasound (AUS), and surface electromyography (sEMG) measurements. The proposed dataset includes lower-limb multimodal data collected from two cohorts, including 30 able-bodied young adults and 12 older adults, across different inclines (0$^\circ$, $\pm$5$^\circ$, and $\pm$10$^\circ$), speeds (0.5 m/s, 1.0 m/s, and 1.5 m/s), and representative non-ideal acquisition conditions (muscle fatigue, electrode shifts, and interday differences). The kinematic and ground reaction force data were collected with a Vicon motion capture system and an instrumented treadmill with embedded force plates, whereas the sEMG and AUS data of thirteen muscles on the bilateral lower limbs were synchronously recorded. K2MUSE is released with the corresponding structured documentation, preprocessing pipelines, and example code, thereby providing a comprehensive resource for rehabilitation robot development, biomechanical analysis, and wearable sensing research. The dataset is available at https://k2muse.github.io/.

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.

Desk Editor's Note private letter to a colleague

K2MUSE is a practical multimodal lower-limb dataset release that adds explicit task and acquisition variability, though its downstream value still needs real checks on synchronization and quality.

read the letter

Hey, I took a look at the K2MUSE paper. The core offering is a new public dataset of lower-limb multimodal walking data that covers both task variations like different speeds and inclines and acquisition variations such as fatigue, electrode shifts, and inter-day differences. It pulls together kinematics and kinetics from Vicon and force plates with sEMG and amplitude mode ultrasound from multiple muscles in 30 young and 12 older adults. What they do well is lay out a clear collection protocol with synchronous recordings and then bundle it with documentation, preprocessing code, and examples. That makes it more than just raw files. The inclusion of older adults and non-ideal conditions is a practical touch that many gait datasets skip, which could help with more realistic model training for rehab robots. The soft spots are fairly minor but worth noting. The paper's value depends on the actual data quality and synchronization, which the abstract describes but the full manuscript would need to demonstrate with specific checks or metrics. Since it's treadmill-based, it might not capture all the variability of real-world walking. The sample sizes are reasonable for this kind of work but not massive, so downstream users will have to test generalization themselves. This is the kind of resource that people in rehabilitation robotics and wearable sensor development would find handy, especially if they're building data-driven systems that need to handle messy real conditions. A reader looking for large-scale multimodal gait samples with built-in variability would get direct value from it. It looks like it deserves a serious referee because it fills a described gap with concrete data and supporting materials. I'd say send it to peer review. Reviewers can confirm the data handling details and help make sure the release is as robust as claimed.

Referee Report

1 major / 2 minor

Summary. The manuscript presents the K2MUSE dataset, a multimodal lower-limb walking collection comprising kinematic data from a Vicon motion capture system, kinetic data from an instrumented treadmill with embedded force plates, surface electromyography (sEMG), and amplitude-mode ultrasound (AUS) from thirteen bilateral lower-limb muscles. Recordings were obtained from two cohorts (30 able-bodied young adults and 12 older adults) across inclines of 0°, ±5°, and ±10°, speeds of 0.5, 1.0, and 1.5 m/s, and non-ideal acquisition conditions including muscle fatigue, electrode shifts, and interday differences. The dataset is released with structured documentation, preprocessing pipelines, and example code at https://k2muse.github.io/.

Significance. If the multimodal signals prove adequately synchronized and representative, the dataset addresses a documented gap in existing lower-limb resources by supplying both task variability and realistic acquisition interference at scale. Inclusion of older-adult data and AUS alongside sEMG supports development of robust, data-driven controllers for rehabilitation robots that must operate under clinical and real-world conditions.

major comments (1)

§3 (Data Collection): the abstract states that kinematic, kinetic, sEMG, and AUS signals 'were synchronously recorded,' yet no quantitative validation of temporal alignment (e.g., cross-correlation latency, hardware trigger description, or maximum inter-modality offset) is supplied; this directly affects usability for time-critical robotic control and must be documented with explicit metrics or procedures.

minor comments (2)

Abstract: the total number of recorded gait cycles or trials per condition is not stated; adding this figure would help readers assess statistical power for downstream machine-learning use cases.
Dataset release: the accompanying documentation should explicitly list the sampling rates of each modality and any applied filtering or synchronization offsets so that users can reproduce the preprocessing pipeline without ambiguity.

Simulated Author's Rebuttal

1 responses · 0 unresolved

We thank the referee for the constructive feedback on our manuscript describing the K2MUSE dataset. We address the single major comment below and will incorporate the requested clarifications in the revised version.

read point-by-point responses

Referee: [—] §3 (Data Collection): the abstract states that kinematic, kinetic, sEMG, and AUS signals 'were synchronously recorded,' yet no quantitative validation of temporal alignment (e.g., cross-correlation latency, hardware trigger description, or maximum inter-modality offset) is supplied; this directly affects usability for time-critical robotic control and must be documented with explicit metrics or procedures.

Authors: We agree that quantitative details on temporal alignment are important for users developing time-critical controllers. Section 3 of the manuscript describes the overall data collection protocol and states that signals were synchronously recorded, but does not provide explicit hardware trigger specifications or measured latency metrics. In the revision we will expand Section 3 to include: (1) a description of the hardware synchronization method (shared trigger signal from the Vicon system to the force-plate treadmill, sEMG amplifier, and AUS acquisition unit), (2) the nominal sampling rates and any known clock offsets, and (3) any post-hoc validation we performed (e.g., cross-correlation of simultaneously recorded events or maximum observed inter-modality offset on representative trials). If additional quantitative measurements can be extracted from the raw acquisition logs, they will be reported; otherwise we will clearly state the synchronization procedure and its known limitations. revision: yes

Circularity Check

0 steps flagged

No significant circularity: dataset release without derivations or predictions

full rationale

The paper is a dataset release describing collection of multimodal lower-limb data (kinematics via Vicon, kinetics via instrumented treadmill, sEMG, and AUS) from able-bodied and older adults under varied inclines, speeds, and non-ideal conditions. No equations, models, predictions, or derivation chains are claimed or present in the abstract or described contribution. The central value rests on public data availability and external downstream use rather than any internal fitting, self-definition, or self-citation that reduces to inputs by construction. This is a standard honest non-finding for data papers.

Axiom & Free-Parameter Ledger

0 free parameters · 2 axioms · 0 invented entities

The contribution rests on standard assumptions of biomechanical measurement accuracy and synchronization rather than new fitted parameters or postulated entities.

axioms (2)

domain assumption Vicon motion capture and instrumented treadmill provide accurate kinematic and kinetic ground truth under the tested conditions.
Invoked in the description of data collection equipment and conditions.
domain assumption Synchronous recording of sEMG and AUS from thirteen bilateral lower-limb muscles is feasible and useful for downstream robot development.
Stated as part of the multimodal collection protocol.

pith-pipeline@v0.9.0 · 5882 in / 1387 out tokens · 60387 ms · 2026-05-22T19:21:28.100102+00:00 · methodology

discussion (0)

Reference graph

Works this paper leans on

17 extracted references · 17 canonical work pages

[1]

Gait & posture 33(1): 6–13

Bovi G, Rabuffetti M, Mazzoleni P and Ferrarin M (2011) A multiple-task gait analysis approach: kinematic, kinetic and emg reference data for healthy young and adult subjects. Gait & posture 33(1): 6–13. Brantley JA, Luu TP , Nakagome S, Zhu F and Contreras-Vidal JL (2018) Full body mobile brain-body imaging data during unconstrained locomotion on stairs,...

work page 2011
[2]

Science robotics 3(15): eaar5438

Ding Y , Kim M, Kuindersma S and Walsh CJ (2018) Human-in-the- loop optimization of hip assistance with a soft exosuit duri ng walking. Science robotics 3(15): eaar5438. Divekar NV , Thomas GC, Yerva AR, Frame HB and Gregg RD (2024) A versatile knee exoskeleton mitigates quadricep s fatigue in lifting, lowering, and carrying tasks. Science Robotics 9(94):...

work page 2018
[3]

IEEE Transactions on Robotics 38(2): 1230–1249

Huo W, Moon H, Alouane MA, Bonnet V , Huang J, Amirat Y , V aidyanathan R and Mohammed S (2021) Impedance modulation control of a lower-limb exoskeleton to assist si t- to-stand movements. IEEE Transactions on Robotics 38(2): 1230–1249. Jin Y , Alvarez JT, Suitor EL, Swaminathan K, Chin A, Civici US, Nuckols RW, Howe RD and Walsh CJ (2024) Estimation of j...

work page 2021
[4]

Gait & posture 54: 325–331

Kainz H, Graham D, Edwards J, Walsh HP , Maine S, Boyd RN, Lloyd DG, Modenese L and Carty CP (2017) Reliability of four models for clinical gait analysis. Gait & posture 54: 325–331. Kang I, Hsu H and Y oung A (2019) The effect of hip assistance levels on human energetic cost using robotic hip exoskeleto ns. IEEE Robotics and Automation Letters 4(2): 430–...

work page 2017
[5]

IEEE Transactions on Neural Systems and Rehabilitation Engineering

Li J, Zhang B, Chen W, Bu C, Zhao Y and Zhao X (2024) Improving hand gesture recognition robustness to dynamic posture variations by multimodal deep feature fusi on. IEEE Transactions on Neural Systems and Rehabilitation Engineering . Liu Z, Han J, Han J and Zhang J (2024) Design and evaluation of a lightweight, ligaments-inspired knee exoskeleton for wa ...

work page 2024
[6]

Science Robotics 9(88): eadi8852

Molinaro DD, Kang I and Y oung AJ (2024a) Estimating human joint moments uniﬁes exoskeleton control, reducing user effort. Science Robotics 9(88): eadi8852. Molinaro DD, Scherpereel KL, Schonhaut EB, Evangelopoulos G, Shepherd MK and Y oung AJ (2024b) Task-agnostic exoskeleton control via biological joint moment estimatio n. Nature 635(8038): 337–344. Mor...

work page 2021
[7]

Gait & posture 7(1): 77–95

Novacheck TF (1998) The biomechanics of running. Gait & posture 7(1): 77–95. Nuckols RW, Lee S, Swaminathan K, Orzel D, Howe RD and Walsh CJ (2021) Individualization of exosuit assistance based on measured muscle dynamics during versatile walking. Science robotics 6(60): eabj1362. Ortiz M, de la Ossa L, Juan J, I´ a˜ nez E, Torricelli D, Tornero J and Azo...

work page 1998
[8]

Scientiﬁc Data 11(1):

Perera CK, Hussain Z, Khant M, Gopalai AA, Gouwanda D and Ahmad SA (2024) A motion capture dataset on human sitting to walking transitions. Scientiﬁc Data 11(1):

work page 2024
[9]

IEEE Transactions on Robotics 34(3): 686–701

Quintero D, Villarreal DJ, Lambert DJ, Kapp S and Gregg RD (2018) Continuous-phase control of a powered knee–ankle prosthesis: Amputee experiments across speeds and incline s. IEEE Transactions on Robotics 34(3): 686–701. Rabuffetti M, Marzegan A, Crippa A, Carpinella I, Lencioni T , Castagna A and Ferrarin M (2019) The lamb gait analysis protocol: Deﬁnit...

work page 2018
[10]

Gait & posture 15(1): 32–44

Riener R, Rabuffetti M and Frigo C (2002) Stair ascent and des cent at different inclinations. Gait & posture 15(1): 32–44. Scherpereel K, Molinaro D, Inan O, Shepherd M and Y oung A (2023) A human lower-limb biomechanics and wearable sensors dataset during cyclic and non-cyclic activities. Scientiﬁc Data 10(1):

work page 2002
[11]

Scientiﬁc Data 10(1):

Schulte RV , Prinsen EC, Schaake L, Paassen RP , Zondag M, van Staveren ES, Poel M and Buurke JH (2023) Database of lower limb kinematics and electromyography during gait- related activities in able-bodied subjects. Scientiﬁc Data 10(1):

work page 2023
[12]

Journal of biomechanics 41(8): 1639–1650

Schwartz MH, Rozumalski A and Trost JP (2008) The effect of walking speed on the gait of typically developing children. Journal of biomechanics 41(8): 1639–1650. Sharma A, Rai V , Calvert M, Dai Z, Guo Z, Boe D and Rombokas E (2023) A non-laboratory gait dataset of full body kinemati cs and egocentric vision. Scientiﬁc Data 10(1):

work page 2008
[13]

Scientiﬁc Data 11(1):

Simonlehner M, Dumphart B and Horsak B (2024) Gaitrec-vr: 3d gait analysis for walking overground with and without a head - mounted-display in virtual reality. Scientiﬁc Data 11(1):

work page 2024
[14]

Nature 610(7931): 277–282

Slade P , Kochenderfer MJ, Delp SL and Collins SH (2022) Personalizing exoskeleton assistance while walking in the real world. Nature 610(7931): 277–282. Tan X, Zhang B, Liu G, Zhao X and Zhao Y (2022) A time- independent control system for natural human gait assistan ce with a soft exoskeleton. IEEE Transactions on Robotics 39(2): 1653–1667. CMU Motion C...

work page 2022
[15]

Science 377(6605): 517–523

Wang C, Chen X, Wang L, Makihata M, Liu HC, Zhou T and Zhao X (2022) Bioadhesive ultrasound for long-term continuous imaging of diverse organs. Science 377(6605): 517–523. Wang G, Jin L, Zhang J, Duan X, Yi J, Zhang M and Sun Z (2024) Recurrent neural network enabled continuous motion estimation of lower limb joints from incomplete semg signal s. IEEE Tra...

work page 2022
[16]

Neural networks 11(7-8): 1317–1329

Wolpert DM and Kawato M (1998) Multiple paired forward and inverse models for motor control. Neural networks 11(7-8): 1317–1329. Wu Q, Xi C, Ding L, Wei C, Ren H, Law R, Dong H and Li XL (2016) Classiﬁcation of emg signals by bfa-optimized gsvcm for diagnosis of fatigue status. IEEE Transactions on Automation Science and Engineering 14(2): 915–930. Yang X...

work page 1998
[17]

IEEE Journal of Biomedical and Health Informatics 26(4): 1718–1725

Zeng J, Zhou Y , Yang Y , Yan J and Liu H (2021) Fatigue- sensitivity comparison of semg and a-mode ultrasound based hand gesture recognition. IEEE Journal of Biomedical and Health Informatics 26(4): 1718–1725. Zhang J, Fiers P , Witte KA, Jackson RW, Poggensee KL, Atkeson CG and Collins SH (2017) Human-in-the-loop optimization of exoskeleton assistance d...

work page 2021

[1] [1]

Gait & posture 33(1): 6–13

Bovi G, Rabuffetti M, Mazzoleni P and Ferrarin M (2011) A multiple-task gait analysis approach: kinematic, kinetic and emg reference data for healthy young and adult subjects. Gait & posture 33(1): 6–13. Brantley JA, Luu TP , Nakagome S, Zhu F and Contreras-Vidal JL (2018) Full body mobile brain-body imaging data during unconstrained locomotion on stairs,...

work page 2011

[2] [2]

Science robotics 3(15): eaar5438

Ding Y , Kim M, Kuindersma S and Walsh CJ (2018) Human-in-the- loop optimization of hip assistance with a soft exosuit duri ng walking. Science robotics 3(15): eaar5438. Divekar NV , Thomas GC, Yerva AR, Frame HB and Gregg RD (2024) A versatile knee exoskeleton mitigates quadricep s fatigue in lifting, lowering, and carrying tasks. Science Robotics 9(94):...

work page 2018

[3] [3]

IEEE Transactions on Robotics 38(2): 1230–1249

Huo W, Moon H, Alouane MA, Bonnet V , Huang J, Amirat Y , V aidyanathan R and Mohammed S (2021) Impedance modulation control of a lower-limb exoskeleton to assist si t- to-stand movements. IEEE Transactions on Robotics 38(2): 1230–1249. Jin Y , Alvarez JT, Suitor EL, Swaminathan K, Chin A, Civici US, Nuckols RW, Howe RD and Walsh CJ (2024) Estimation of j...

work page 2021

[4] [4]

Gait & posture 54: 325–331

Kainz H, Graham D, Edwards J, Walsh HP , Maine S, Boyd RN, Lloyd DG, Modenese L and Carty CP (2017) Reliability of four models for clinical gait analysis. Gait & posture 54: 325–331. Kang I, Hsu H and Y oung A (2019) The effect of hip assistance levels on human energetic cost using robotic hip exoskeleto ns. IEEE Robotics and Automation Letters 4(2): 430–...

work page 2017

[5] [5]

IEEE Transactions on Neural Systems and Rehabilitation Engineering

Li J, Zhang B, Chen W, Bu C, Zhao Y and Zhao X (2024) Improving hand gesture recognition robustness to dynamic posture variations by multimodal deep feature fusi on. IEEE Transactions on Neural Systems and Rehabilitation Engineering . Liu Z, Han J, Han J and Zhang J (2024) Design and evaluation of a lightweight, ligaments-inspired knee exoskeleton for wa ...

work page 2024

[6] [6]

Science Robotics 9(88): eadi8852

Molinaro DD, Kang I and Y oung AJ (2024a) Estimating human joint moments uniﬁes exoskeleton control, reducing user effort. Science Robotics 9(88): eadi8852. Molinaro DD, Scherpereel KL, Schonhaut EB, Evangelopoulos G, Shepherd MK and Y oung AJ (2024b) Task-agnostic exoskeleton control via biological joint moment estimatio n. Nature 635(8038): 337–344. Mor...

work page 2021

[7] [7]

Gait & posture 7(1): 77–95

Novacheck TF (1998) The biomechanics of running. Gait & posture 7(1): 77–95. Nuckols RW, Lee S, Swaminathan K, Orzel D, Howe RD and Walsh CJ (2021) Individualization of exosuit assistance based on measured muscle dynamics during versatile walking. Science robotics 6(60): eabj1362. Ortiz M, de la Ossa L, Juan J, I´ a˜ nez E, Torricelli D, Tornero J and Azo...

work page 1998

[8] [8]

Scientiﬁc Data 11(1):

Perera CK, Hussain Z, Khant M, Gopalai AA, Gouwanda D and Ahmad SA (2024) A motion capture dataset on human sitting to walking transitions. Scientiﬁc Data 11(1):

work page 2024

[9] [9]

IEEE Transactions on Robotics 34(3): 686–701

Quintero D, Villarreal DJ, Lambert DJ, Kapp S and Gregg RD (2018) Continuous-phase control of a powered knee–ankle prosthesis: Amputee experiments across speeds and incline s. IEEE Transactions on Robotics 34(3): 686–701. Rabuffetti M, Marzegan A, Crippa A, Carpinella I, Lencioni T , Castagna A and Ferrarin M (2019) The lamb gait analysis protocol: Deﬁnit...

work page 2018

[10] [10]

Gait & posture 15(1): 32–44

Riener R, Rabuffetti M and Frigo C (2002) Stair ascent and des cent at different inclinations. Gait & posture 15(1): 32–44. Scherpereel K, Molinaro D, Inan O, Shepherd M and Y oung A (2023) A human lower-limb biomechanics and wearable sensors dataset during cyclic and non-cyclic activities. Scientiﬁc Data 10(1):

work page 2002

[11] [11]

Scientiﬁc Data 10(1):

Schulte RV , Prinsen EC, Schaake L, Paassen RP , Zondag M, van Staveren ES, Poel M and Buurke JH (2023) Database of lower limb kinematics and electromyography during gait- related activities in able-bodied subjects. Scientiﬁc Data 10(1):

work page 2023

[12] [12]

Journal of biomechanics 41(8): 1639–1650

Schwartz MH, Rozumalski A and Trost JP (2008) The effect of walking speed on the gait of typically developing children. Journal of biomechanics 41(8): 1639–1650. Sharma A, Rai V , Calvert M, Dai Z, Guo Z, Boe D and Rombokas E (2023) A non-laboratory gait dataset of full body kinemati cs and egocentric vision. Scientiﬁc Data 10(1):

work page 2008

[13] [13]

Scientiﬁc Data 11(1):

Simonlehner M, Dumphart B and Horsak B (2024) Gaitrec-vr: 3d gait analysis for walking overground with and without a head - mounted-display in virtual reality. Scientiﬁc Data 11(1):

work page 2024

[14] [14]

Nature 610(7931): 277–282

Slade P , Kochenderfer MJ, Delp SL and Collins SH (2022) Personalizing exoskeleton assistance while walking in the real world. Nature 610(7931): 277–282. Tan X, Zhang B, Liu G, Zhao X and Zhao Y (2022) A time- independent control system for natural human gait assistan ce with a soft exoskeleton. IEEE Transactions on Robotics 39(2): 1653–1667. CMU Motion C...

work page 2022

[15] [15]

Science 377(6605): 517–523

Wang C, Chen X, Wang L, Makihata M, Liu HC, Zhou T and Zhao X (2022) Bioadhesive ultrasound for long-term continuous imaging of diverse organs. Science 377(6605): 517–523. Wang G, Jin L, Zhang J, Duan X, Yi J, Zhang M and Sun Z (2024) Recurrent neural network enabled continuous motion estimation of lower limb joints from incomplete semg signal s. IEEE Tra...

work page 2022

[16] [16]

Neural networks 11(7-8): 1317–1329

Wolpert DM and Kawato M (1998) Multiple paired forward and inverse models for motor control. Neural networks 11(7-8): 1317–1329. Wu Q, Xi C, Ding L, Wei C, Ren H, Law R, Dong H and Li XL (2016) Classiﬁcation of emg signals by bfa-optimized gsvcm for diagnosis of fatigue status. IEEE Transactions on Automation Science and Engineering 14(2): 915–930. Yang X...

work page 1998

[17] [17]

IEEE Journal of Biomedical and Health Informatics 26(4): 1718–1725

Zeng J, Zhou Y , Yang Y , Yan J and Liu H (2021) Fatigue- sensitivity comparison of semg and a-mode ultrasound based hand gesture recognition. IEEE Journal of Biomedical and Health Informatics 26(4): 1718–1725. Zhang J, Fiers P , Witte KA, Jackson RW, Poggensee KL, Atkeson CG and Collins SH (2017) Human-in-the-loop optimization of exoskeleton assistance d...

work page 2021