Learning based lower limb joint kinematic estimation using open source IMU data

Abstract This study introduces a deep learning framework for estimating lower-limb joint kinematics using inertial measurement units (IMUs). While deep learning methods avoid sensor drift, extensive calibration, and complex setup procedures, they require substantial data. To meet this demand, we lev...

Full description

Saved in:
Bibliographic Details
Main Authors: Benjamin Hur, Sunin Baek, Inseung Kang, Daekyum Kim
Format: Article
Language:English
Published: Nature Portfolio 2025-02-01
Series:Scientific Reports
Online Access:https://doi.org/10.1038/s41598-025-89716-4
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1850198041486688256
author Benjamin Hur
Sunin Baek
Inseung Kang
Daekyum Kim
author_facet Benjamin Hur
Sunin Baek
Inseung Kang
Daekyum Kim
author_sort Benjamin Hur
collection DOAJ
description Abstract This study introduces a deep learning framework for estimating lower-limb joint kinematics using inertial measurement units (IMUs). While deep learning methods avoid sensor drift, extensive calibration, and complex setup procedures, they require substantial data. To meet this demand, we leveraged an open-source dataset to develop and evaluate three training approaches. The first involved training a model exclusively on data from a single user, resulting in high accuracy for that individual only. The second approach trained a model on data from multiple users to generalize across individuals; however, demonstrated lower accuracy due to variations in gait patterns across users. The third approach added transfer learning to the second, improving estimation accuracy for new users through fine-tuning with a small portion of their data. This model overcame the limitations of the previous methods’ dependency on extensive data collection, and achieved comparable performance to inverse kinematics, making it an effective solution for diverse populations. Additionally, our analysis on IMU combinations suggests that IMUs placed on the femur and calcaneus are the best for most cases. This framework not only reduces the need for extensive data collection but also enhances personalized gait analysis, enabling more efficient and accessible applications in both clinical assessments and real-world environments for broader use.
format Article
id doaj-art-1534608d3e8647749abea72c38cc4eee
institution OA Journals
issn 2045-2322
language English
publishDate 2025-02-01
publisher Nature Portfolio
record_format Article
series Scientific Reports
spelling doaj-art-1534608d3e8647749abea72c38cc4eee2025-08-20T02:12:58ZengNature PortfolioScientific Reports2045-23222025-02-0115111210.1038/s41598-025-89716-4Learning based lower limb joint kinematic estimation using open source IMU dataBenjamin Hur0Sunin Baek1Inseung Kang2Daekyum Kim3Korea University, School of Mechanical EngineeringKorea University, School of Mechanical EngineeringDepartment of Mechanical Engineering, Carnegie Mellon UniversityKorea University, School of Mechanical EngineeringAbstract This study introduces a deep learning framework for estimating lower-limb joint kinematics using inertial measurement units (IMUs). While deep learning methods avoid sensor drift, extensive calibration, and complex setup procedures, they require substantial data. To meet this demand, we leveraged an open-source dataset to develop and evaluate three training approaches. The first involved training a model exclusively on data from a single user, resulting in high accuracy for that individual only. The second approach trained a model on data from multiple users to generalize across individuals; however, demonstrated lower accuracy due to variations in gait patterns across users. The third approach added transfer learning to the second, improving estimation accuracy for new users through fine-tuning with a small portion of their data. This model overcame the limitations of the previous methods’ dependency on extensive data collection, and achieved comparable performance to inverse kinematics, making it an effective solution for diverse populations. Additionally, our analysis on IMU combinations suggests that IMUs placed on the femur and calcaneus are the best for most cases. This framework not only reduces the need for extensive data collection but also enhances personalized gait analysis, enabling more efficient and accessible applications in both clinical assessments and real-world environments for broader use.https://doi.org/10.1038/s41598-025-89716-4
spellingShingle Benjamin Hur
Sunin Baek
Inseung Kang
Daekyum Kim
Learning based lower limb joint kinematic estimation using open source IMU data
Scientific Reports
title Learning based lower limb joint kinematic estimation using open source IMU data
title_full Learning based lower limb joint kinematic estimation using open source IMU data
title_fullStr Learning based lower limb joint kinematic estimation using open source IMU data
title_full_unstemmed Learning based lower limb joint kinematic estimation using open source IMU data
title_short Learning based lower limb joint kinematic estimation using open source IMU data
title_sort learning based lower limb joint kinematic estimation using open source imu data
url https://doi.org/10.1038/s41598-025-89716-4
work_keys_str_mv AT benjaminhur learningbasedlowerlimbjointkinematicestimationusingopensourceimudata
AT suninbaek learningbasedlowerlimbjointkinematicestimationusingopensourceimudata
AT inseungkang learningbasedlowerlimbjointkinematicestimationusingopensourceimudata
AT daekyumkim learningbasedlowerlimbjointkinematicestimationusingopensourceimudata