Exploiting motion correlations in 3-D articulated human motion tracking

Xinyu Xu, Baoxin Li

Research output: Contribution to journalArticle

13 Citations (Scopus)

Abstract

In 3-D articulated human motion tracking, the curse of dimensionality renders commonly-used particle-filter-based approaches inefficient. Also, noisy image measurements and imperfect feature extraction call for strong motion prior. We propose to learn the correlation between the right-side and the left-side human motion using partial least square (PLS) regression. The correlation effectively constrains the sampling of the proposal distribution to portions of the parameter space that correspond to plausible human motions. The learned correlation is then used as motion prior in designing a Rao-Blackwellized particle filter algorithm, RBPF-PLS, which estimates only one group of state variables using the Monte Carlo method, leaving the other group being exactly computed through an analytical filter that utilizes the learned motion correlation. We quantitatively assessed the accuracy of the proposed algorithm with challenging HumanEva-I/II data set. Experiments with comparison with both the annealed particle filter and the standard particle filter show that the proposed method achieves lower estimation error in processing challenging real-world data of 3-D human motion. In particular, the experiments demonstrate that the learned motion correlation model generalizes well to motions outside of the training set and is insensitive to the choice of the training subjects, suggesting the potential wide applicability of the method.

Original languageEnglish (US)
Pages (from-to)1292-1303
Number of pages12
JournalIEEE Transactions on Image Processing
Volume18
Issue number6
DOIs
StatePublished - 2009

Fingerprint

Error analysis
Feature extraction
Monte Carlo methods
Experiments
Sampling
Processing

Keywords

  • 3-D articulated human motion tracking
  • Partial least square regression
  • Particle filtering
  • Rao-Blackwellized particle filter (RBPF)

ASJC Scopus subject areas

  • Computer Graphics and Computer-Aided Design
  • Software

Cite this

Exploiting motion correlations in 3-D articulated human motion tracking. / Xu, Xinyu; Li, Baoxin.

In: IEEE Transactions on Image Processing, Vol. 18, No. 6, 2009, p. 1292-1303.

Research output: Contribution to journalArticle

@article{e778355daf5d43cd989a5653d7fa7ec0,
title = "Exploiting motion correlations in 3-D articulated human motion tracking",
abstract = "In 3-D articulated human motion tracking, the curse of dimensionality renders commonly-used particle-filter-based approaches inefficient. Also, noisy image measurements and imperfect feature extraction call for strong motion prior. We propose to learn the correlation between the right-side and the left-side human motion using partial least square (PLS) regression. The correlation effectively constrains the sampling of the proposal distribution to portions of the parameter space that correspond to plausible human motions. The learned correlation is then used as motion prior in designing a Rao-Blackwellized particle filter algorithm, RBPF-PLS, which estimates only one group of state variables using the Monte Carlo method, leaving the other group being exactly computed through an analytical filter that utilizes the learned motion correlation. We quantitatively assessed the accuracy of the proposed algorithm with challenging HumanEva-I/II data set. Experiments with comparison with both the annealed particle filter and the standard particle filter show that the proposed method achieves lower estimation error in processing challenging real-world data of 3-D human motion. In particular, the experiments demonstrate that the learned motion correlation model generalizes well to motions outside of the training set and is insensitive to the choice of the training subjects, suggesting the potential wide applicability of the method.",
keywords = "3-D articulated human motion tracking, Partial least square regression, Particle filtering, Rao-Blackwellized particle filter (RBPF)",
author = "Xinyu Xu and Baoxin Li",
year = "2009",
doi = "10.1109/TIP.2009.2017131",
language = "English (US)",
volume = "18",
pages = "1292--1303",
journal = "IEEE Transactions on Image Processing",
issn = "1057-7149",
publisher = "Institute of Electrical and Electronics Engineers Inc.",
number = "6",

}

TY - JOUR

T1 - Exploiting motion correlations in 3-D articulated human motion tracking

AU - Xu, Xinyu

AU - Li, Baoxin

PY - 2009

Y1 - 2009

N2 - In 3-D articulated human motion tracking, the curse of dimensionality renders commonly-used particle-filter-based approaches inefficient. Also, noisy image measurements and imperfect feature extraction call for strong motion prior. We propose to learn the correlation between the right-side and the left-side human motion using partial least square (PLS) regression. The correlation effectively constrains the sampling of the proposal distribution to portions of the parameter space that correspond to plausible human motions. The learned correlation is then used as motion prior in designing a Rao-Blackwellized particle filter algorithm, RBPF-PLS, which estimates only one group of state variables using the Monte Carlo method, leaving the other group being exactly computed through an analytical filter that utilizes the learned motion correlation. We quantitatively assessed the accuracy of the proposed algorithm with challenging HumanEva-I/II data set. Experiments with comparison with both the annealed particle filter and the standard particle filter show that the proposed method achieves lower estimation error in processing challenging real-world data of 3-D human motion. In particular, the experiments demonstrate that the learned motion correlation model generalizes well to motions outside of the training set and is insensitive to the choice of the training subjects, suggesting the potential wide applicability of the method.

AB - In 3-D articulated human motion tracking, the curse of dimensionality renders commonly-used particle-filter-based approaches inefficient. Also, noisy image measurements and imperfect feature extraction call for strong motion prior. We propose to learn the correlation between the right-side and the left-side human motion using partial least square (PLS) regression. The correlation effectively constrains the sampling of the proposal distribution to portions of the parameter space that correspond to plausible human motions. The learned correlation is then used as motion prior in designing a Rao-Blackwellized particle filter algorithm, RBPF-PLS, which estimates only one group of state variables using the Monte Carlo method, leaving the other group being exactly computed through an analytical filter that utilizes the learned motion correlation. We quantitatively assessed the accuracy of the proposed algorithm with challenging HumanEva-I/II data set. Experiments with comparison with both the annealed particle filter and the standard particle filter show that the proposed method achieves lower estimation error in processing challenging real-world data of 3-D human motion. In particular, the experiments demonstrate that the learned motion correlation model generalizes well to motions outside of the training set and is insensitive to the choice of the training subjects, suggesting the potential wide applicability of the method.

KW - 3-D articulated human motion tracking

KW - Partial least square regression

KW - Particle filtering

KW - Rao-Blackwellized particle filter (RBPF)

UR - http://www.scopus.com/inward/record.url?scp=67349101860&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=67349101860&partnerID=8YFLogxK

U2 - 10.1109/TIP.2009.2017131

DO - 10.1109/TIP.2009.2017131

M3 - Article

C2 - 19369156

AN - SCOPUS:67349101860

VL - 18

SP - 1292

EP - 1303

JO - IEEE Transactions on Image Processing

JF - IEEE Transactions on Image Processing

SN - 1057-7149

IS - 6

ER -