Gaussian Process Trajectory Learning and Synthesis of Individualized Gait Motions

Jisoo Hong, Changmook Chun, Seung-Jong Kim, Frank C. Park

Research output: Contribution to journalArticle

Abstract

This paper proposes a Gaussian process-based method for trajectory learning and generation of individualized gait motions at arbitrary user-designated walking speeds, intended to be used in generating reference motions for robotic gait rehabilitation systems. We utilize a nonlinear dimension reduction technique based on Gaussian process dynamical models (GPDMs), in which the internal dynamics is modeled as a second-order Markov process evolving in a lower-dimensional latent space. After the GPDM parameters are identified with training data obtained from gait motions of healthy subjects walking at different speeds, our method then employs Gaussian process regression (GPR) to predict the initial two states of the latent space dynamics from any arbitrary desired walking speed and the anthropometric parameters of the test subject. Motions are then generated by directly mapping the latent space dynamics to joint trajectories. Experimental studies involving more than 100 subjects indicate that our method generates gait patterns with 30% less mean square prediction errors compared to recent state-of-the-art methods, while also allowing for arbitrary user-specified walking speeds.

Original languageEnglish
Article number8703438
Pages (from-to)1236-1245
Number of pages10
JournalIEEE Transactions on Neural Systems and Rehabilitation Engineering
Volume27
Issue number6
DOIs
Publication statusPublished - 2019 Jun 1

Fingerprint

Gait
Trajectories
Learning
Markov Chains
Robotics
Patient rehabilitation
Markov processes
Walking
Healthy Volunteers
Rehabilitation
Joints
Walking Speed

Keywords

  • Gait rehabilitation
  • Gaussian process dynamical model
  • Gaussian process regression
  • robot rehabilitation

ASJC Scopus subject areas

  • Neuroscience(all)
  • Biomedical Engineering
  • Computer Science Applications

Cite this

Gaussian Process Trajectory Learning and Synthesis of Individualized Gait Motions. / Hong, Jisoo; Chun, Changmook; Kim, Seung-Jong; Park, Frank C.

In: IEEE Transactions on Neural Systems and Rehabilitation Engineering, Vol. 27, No. 6, 8703438, 01.06.2019, p. 1236-1245.

Research output: Contribution to journalArticle

@article{b8f0ab7437e947ad8a14aac0006e30b8,
title = "Gaussian Process Trajectory Learning and Synthesis of Individualized Gait Motions",
abstract = "This paper proposes a Gaussian process-based method for trajectory learning and generation of individualized gait motions at arbitrary user-designated walking speeds, intended to be used in generating reference motions for robotic gait rehabilitation systems. We utilize a nonlinear dimension reduction technique based on Gaussian process dynamical models (GPDMs), in which the internal dynamics is modeled as a second-order Markov process evolving in a lower-dimensional latent space. After the GPDM parameters are identified with training data obtained from gait motions of healthy subjects walking at different speeds, our method then employs Gaussian process regression (GPR) to predict the initial two states of the latent space dynamics from any arbitrary desired walking speed and the anthropometric parameters of the test subject. Motions are then generated by directly mapping the latent space dynamics to joint trajectories. Experimental studies involving more than 100 subjects indicate that our method generates gait patterns with 30{\%} less mean square prediction errors compared to recent state-of-the-art methods, while also allowing for arbitrary user-specified walking speeds.",
keywords = "Gait rehabilitation, Gaussian process dynamical model, Gaussian process regression, robot rehabilitation",
author = "Jisoo Hong and Changmook Chun and Seung-Jong Kim and Park, {Frank C.}",
year = "2019",
month = "6",
day = "1",
doi = "10.1109/TNSRE.2019.2914095",
language = "English",
volume = "27",
pages = "1236--1245",
journal = "IEEE Transactions on Neural Systems and Rehabilitation Engineering",
issn = "1534-4320",
publisher = "Institute of Electrical and Electronics Engineers Inc.",
number = "6",

}

TY - JOUR

T1 - Gaussian Process Trajectory Learning and Synthesis of Individualized Gait Motions

AU - Hong, Jisoo

AU - Chun, Changmook

AU - Kim, Seung-Jong

AU - Park, Frank C.

PY - 2019/6/1

Y1 - 2019/6/1

N2 - This paper proposes a Gaussian process-based method for trajectory learning and generation of individualized gait motions at arbitrary user-designated walking speeds, intended to be used in generating reference motions for robotic gait rehabilitation systems. We utilize a nonlinear dimension reduction technique based on Gaussian process dynamical models (GPDMs), in which the internal dynamics is modeled as a second-order Markov process evolving in a lower-dimensional latent space. After the GPDM parameters are identified with training data obtained from gait motions of healthy subjects walking at different speeds, our method then employs Gaussian process regression (GPR) to predict the initial two states of the latent space dynamics from any arbitrary desired walking speed and the anthropometric parameters of the test subject. Motions are then generated by directly mapping the latent space dynamics to joint trajectories. Experimental studies involving more than 100 subjects indicate that our method generates gait patterns with 30% less mean square prediction errors compared to recent state-of-the-art methods, while also allowing for arbitrary user-specified walking speeds.

AB - This paper proposes a Gaussian process-based method for trajectory learning and generation of individualized gait motions at arbitrary user-designated walking speeds, intended to be used in generating reference motions for robotic gait rehabilitation systems. We utilize a nonlinear dimension reduction technique based on Gaussian process dynamical models (GPDMs), in which the internal dynamics is modeled as a second-order Markov process evolving in a lower-dimensional latent space. After the GPDM parameters are identified with training data obtained from gait motions of healthy subjects walking at different speeds, our method then employs Gaussian process regression (GPR) to predict the initial two states of the latent space dynamics from any arbitrary desired walking speed and the anthropometric parameters of the test subject. Motions are then generated by directly mapping the latent space dynamics to joint trajectories. Experimental studies involving more than 100 subjects indicate that our method generates gait patterns with 30% less mean square prediction errors compared to recent state-of-the-art methods, while also allowing for arbitrary user-specified walking speeds.

KW - Gait rehabilitation

KW - Gaussian process dynamical model

KW - Gaussian process regression

KW - robot rehabilitation

UR - http://www.scopus.com/inward/record.url?scp=85067261594&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85067261594&partnerID=8YFLogxK

U2 - 10.1109/TNSRE.2019.2914095

DO - 10.1109/TNSRE.2019.2914095

M3 - Article

VL - 27

SP - 1236

EP - 1245

JO - IEEE Transactions on Neural Systems and Rehabilitation Engineering

JF - IEEE Transactions on Neural Systems and Rehabilitation Engineering

SN - 1534-4320

IS - 6

M1 - 8703438

ER -