Learning-based 3D human kinematics estimation using behavioral constraints from activity classification

Abstract Inertial measurement units offer a cost-effective, portable alternative to lab-based motion capture systems. However, measuring joint angles and movement trajectories with inertial measurement units is challenging due to signal drift errors caused by biases and noise, which are amplified by...

Full description

Saved in:
Bibliographic Details
Main Authors: Daekyum Kim, Yichu Jin, Haedo Cho, Truman Jones, Yu Meng Zhou, Ameneh Fadaie, Dmitry Popov, Krithika Swaminathan, Conor J. Walsh
Format: Article
Language:English
Published: Nature Portfolio 2025-04-01
Series:Nature Communications
Online Access:https://doi.org/10.1038/s41467-025-58624-6
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1849737601496383488
author Daekyum Kim
Yichu Jin
Haedo Cho
Truman Jones
Yu Meng Zhou
Ameneh Fadaie
Dmitry Popov
Krithika Swaminathan
Conor J. Walsh
author_facet Daekyum Kim
Yichu Jin
Haedo Cho
Truman Jones
Yu Meng Zhou
Ameneh Fadaie
Dmitry Popov
Krithika Swaminathan
Conor J. Walsh
author_sort Daekyum Kim
collection DOAJ
description Abstract Inertial measurement units offer a cost-effective, portable alternative to lab-based motion capture systems. However, measuring joint angles and movement trajectories with inertial measurement units is challenging due to signal drift errors caused by biases and noise, which are amplified by numerical integration. Existing approaches use anatomical constraints to reduce drift but require body parameter measurements. Learning-based approaches show promise but often lack accuracy for broad applications (e.g., strength training). Here, we introduce the Activity-in-the-loop Kinematics Estimator, an end-to-end machine learning model incorporating human behavioral constraints for enhanced kinematics estimation using two inertial measurement units. It integrates activity classification with kinematics estimation, leveraging limited movement patterns during specific activities. In dynamic scenarios, our approach achieved trajectory and shoulder joint angle errors under 0.021 m and $$6.5^\circ$$ 6 . 5 ∘ , respectively, 52% and 17% lower than errors without including activity classification. These results highlight accurate motion tracking with minimal inertial measurement units and domain-specific context.
format Article
id doaj-art-b8b80f612fac496aa7c01839cc30d35a
institution DOAJ
issn 2041-1723
language English
publishDate 2025-04-01
publisher Nature Portfolio
record_format Article
series Nature Communications
spelling doaj-art-b8b80f612fac496aa7c01839cc30d35a2025-08-20T03:06:52ZengNature PortfolioNature Communications2041-17232025-04-0116111110.1038/s41467-025-58624-6Learning-based 3D human kinematics estimation using behavioral constraints from activity classificationDaekyum Kim0Yichu Jin1Haedo Cho2Truman Jones3Yu Meng Zhou4Ameneh Fadaie5Dmitry Popov6Krithika Swaminathan7Conor J. Walsh8John A. Paulson School of Engineering and Applied Sciences, Harvard UniversityJohn A. Paulson School of Engineering and Applied Sciences, Harvard UniversityJohn A. Paulson School of Engineering and Applied Sciences, Harvard UniversityJohn A. Paulson School of Engineering and Applied Sciences, Harvard UniversityJohn A. Paulson School of Engineering and Applied Sciences, Harvard UniversityJohn A. Paulson School of Engineering and Applied Sciences, Harvard UniversityJohn A. Paulson School of Engineering and Applied Sciences, Harvard UniversityJohn A. Paulson School of Engineering and Applied Sciences, Harvard UniversityJohn A. Paulson School of Engineering and Applied Sciences, Harvard UniversityAbstract Inertial measurement units offer a cost-effective, portable alternative to lab-based motion capture systems. However, measuring joint angles and movement trajectories with inertial measurement units is challenging due to signal drift errors caused by biases and noise, which are amplified by numerical integration. Existing approaches use anatomical constraints to reduce drift but require body parameter measurements. Learning-based approaches show promise but often lack accuracy for broad applications (e.g., strength training). Here, we introduce the Activity-in-the-loop Kinematics Estimator, an end-to-end machine learning model incorporating human behavioral constraints for enhanced kinematics estimation using two inertial measurement units. It integrates activity classification with kinematics estimation, leveraging limited movement patterns during specific activities. In dynamic scenarios, our approach achieved trajectory and shoulder joint angle errors under 0.021 m and $$6.5^\circ$$ 6 . 5 ∘ , respectively, 52% and 17% lower than errors without including activity classification. These results highlight accurate motion tracking with minimal inertial measurement units and domain-specific context.https://doi.org/10.1038/s41467-025-58624-6
spellingShingle Daekyum Kim
Yichu Jin
Haedo Cho
Truman Jones
Yu Meng Zhou
Ameneh Fadaie
Dmitry Popov
Krithika Swaminathan
Conor J. Walsh
Learning-based 3D human kinematics estimation using behavioral constraints from activity classification
Nature Communications
title Learning-based 3D human kinematics estimation using behavioral constraints from activity classification
title_full Learning-based 3D human kinematics estimation using behavioral constraints from activity classification
title_fullStr Learning-based 3D human kinematics estimation using behavioral constraints from activity classification
title_full_unstemmed Learning-based 3D human kinematics estimation using behavioral constraints from activity classification
title_short Learning-based 3D human kinematics estimation using behavioral constraints from activity classification
title_sort learning based 3d human kinematics estimation using behavioral constraints from activity classification
url https://doi.org/10.1038/s41467-025-58624-6
work_keys_str_mv AT daekyumkim learningbased3dhumankinematicsestimationusingbehavioralconstraintsfromactivityclassification
AT yichujin learningbased3dhumankinematicsestimationusingbehavioralconstraintsfromactivityclassification
AT haedocho learningbased3dhumankinematicsestimationusingbehavioralconstraintsfromactivityclassification
AT trumanjones learningbased3dhumankinematicsestimationusingbehavioralconstraintsfromactivityclassification
AT yumengzhou learningbased3dhumankinematicsestimationusingbehavioralconstraintsfromactivityclassification
AT amenehfadaie learningbased3dhumankinematicsestimationusingbehavioralconstraintsfromactivityclassification
AT dmitrypopov learningbased3dhumankinematicsestimationusingbehavioralconstraintsfromactivityclassification
AT krithikaswaminathan learningbased3dhumankinematicsestimationusingbehavioralconstraintsfromactivityclassification
AT conorjwalsh learningbased3dhumankinematicsestimationusingbehavioralconstraintsfromactivityclassification