Unsupervised Temporal Segmentation of Repetitive Human Actions Based on Kinematic Modeling and Frequency Analysis

Qifei Wang, Gregorij Kurillo, Ferda Ofli, Ruzena Bajcsy

Research output: Chapter in Book/Report/Conference proceedingConference contribution

13 Citations (Scopus)

Abstract

In this paper, we propose a method for temporal segmentation of human repetitive actions based on frequency analysis of kinematic parameters, zero-velocity crossing detection, and adaptive k-means clustering. Since the human motion data may be captured with different modalities which have different temporal sampling rate and accuracy (e.g., Optical motion capture systems vs. Microsoft Kinect), we first apply a generic full-body kinematic model with an unscented Kalman filter to convert the motion data into a unified representation that is robust to noise. Furthermore, we extract the most representative kinematic parameters via the primary frequency analysis. The sequences are segmented based on zero-velocity crossing of the selected parameters followed by an adaptive k-means clustering to identify the repetition segments. Experimental results demonstrate that for the motion data captured by both the motion capture system and the Microsoft Kinect, our proposed algorithm obtains robust segmentation of repetitive action sequences.

Original languageEnglish
Title of host publicationProceedings - 2015 International Conference on 3D Vision, 3DV 2015
PublisherInstitute of Electrical and Electronics Engineers Inc.
Pages562-570
Number of pages9
ISBN (Electronic)9781467383325
DOIs
Publication statusPublished - 20 Nov 2015
Externally publishedYes
Event2015 International Conference on 3D Vision, 3DV 2015 - Lyon, France
Duration: 19 Oct 201522 Oct 2015

Other

Other2015 International Conference on 3D Vision, 3DV 2015
CountryFrance
CityLyon
Period19/10/1522/10/15

Fingerprint

Kinematics
Kalman filters
Sampling

Keywords

  • Bones
  • Computer vision
  • Extremities
  • Hidden Markov models
  • Joints
  • Kinematics
  • Motion segmentation

ASJC Scopus subject areas

  • Computer Vision and Pattern Recognition

Cite this

Wang, Q., Kurillo, G., Ofli, F., & Bajcsy, R. (2015). Unsupervised Temporal Segmentation of Repetitive Human Actions Based on Kinematic Modeling and Frequency Analysis. In Proceedings - 2015 International Conference on 3D Vision, 3DV 2015 (pp. 562-570). [7335526] Institute of Electrical and Electronics Engineers Inc.. https://doi.org/10.1109/3DV.2015.69

Unsupervised Temporal Segmentation of Repetitive Human Actions Based on Kinematic Modeling and Frequency Analysis. / Wang, Qifei; Kurillo, Gregorij; Ofli, Ferda; Bajcsy, Ruzena.

Proceedings - 2015 International Conference on 3D Vision, 3DV 2015. Institute of Electrical and Electronics Engineers Inc., 2015. p. 562-570 7335526.

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Wang, Q, Kurillo, G, Ofli, F & Bajcsy, R 2015, Unsupervised Temporal Segmentation of Repetitive Human Actions Based on Kinematic Modeling and Frequency Analysis. in Proceedings - 2015 International Conference on 3D Vision, 3DV 2015., 7335526, Institute of Electrical and Electronics Engineers Inc., pp. 562-570, 2015 International Conference on 3D Vision, 3DV 2015, Lyon, France, 19/10/15. https://doi.org/10.1109/3DV.2015.69
Wang Q, Kurillo G, Ofli F, Bajcsy R. Unsupervised Temporal Segmentation of Repetitive Human Actions Based on Kinematic Modeling and Frequency Analysis. In Proceedings - 2015 International Conference on 3D Vision, 3DV 2015. Institute of Electrical and Electronics Engineers Inc. 2015. p. 562-570. 7335526 https://doi.org/10.1109/3DV.2015.69
Wang, Qifei ; Kurillo, Gregorij ; Ofli, Ferda ; Bajcsy, Ruzena. / Unsupervised Temporal Segmentation of Repetitive Human Actions Based on Kinematic Modeling and Frequency Analysis. Proceedings - 2015 International Conference on 3D Vision, 3DV 2015. Institute of Electrical and Electronics Engineers Inc., 2015. pp. 562-570
@inproceedings{bfada97ed2a3446c9a6b5a96a22b4c23,
title = "Unsupervised Temporal Segmentation of Repetitive Human Actions Based on Kinematic Modeling and Frequency Analysis",
abstract = "In this paper, we propose a method for temporal segmentation of human repetitive actions based on frequency analysis of kinematic parameters, zero-velocity crossing detection, and adaptive k-means clustering. Since the human motion data may be captured with different modalities which have different temporal sampling rate and accuracy (e.g., Optical motion capture systems vs. Microsoft Kinect), we first apply a generic full-body kinematic model with an unscented Kalman filter to convert the motion data into a unified representation that is robust to noise. Furthermore, we extract the most representative kinematic parameters via the primary frequency analysis. The sequences are segmented based on zero-velocity crossing of the selected parameters followed by an adaptive k-means clustering to identify the repetition segments. Experimental results demonstrate that for the motion data captured by both the motion capture system and the Microsoft Kinect, our proposed algorithm obtains robust segmentation of repetitive action sequences.",
keywords = "Bones, Computer vision, Extremities, Hidden Markov models, Joints, Kinematics, Motion segmentation",
author = "Qifei Wang and Gregorij Kurillo and Ferda Ofli and Ruzena Bajcsy",
year = "2015",
month = "11",
day = "20",
doi = "10.1109/3DV.2015.69",
language = "English",
pages = "562--570",
booktitle = "Proceedings - 2015 International Conference on 3D Vision, 3DV 2015",
publisher = "Institute of Electrical and Electronics Engineers Inc.",

}

TY - GEN

T1 - Unsupervised Temporal Segmentation of Repetitive Human Actions Based on Kinematic Modeling and Frequency Analysis

AU - Wang, Qifei

AU - Kurillo, Gregorij

AU - Ofli, Ferda

AU - Bajcsy, Ruzena

PY - 2015/11/20

Y1 - 2015/11/20

N2 - In this paper, we propose a method for temporal segmentation of human repetitive actions based on frequency analysis of kinematic parameters, zero-velocity crossing detection, and adaptive k-means clustering. Since the human motion data may be captured with different modalities which have different temporal sampling rate and accuracy (e.g., Optical motion capture systems vs. Microsoft Kinect), we first apply a generic full-body kinematic model with an unscented Kalman filter to convert the motion data into a unified representation that is robust to noise. Furthermore, we extract the most representative kinematic parameters via the primary frequency analysis. The sequences are segmented based on zero-velocity crossing of the selected parameters followed by an adaptive k-means clustering to identify the repetition segments. Experimental results demonstrate that for the motion data captured by both the motion capture system and the Microsoft Kinect, our proposed algorithm obtains robust segmentation of repetitive action sequences.

AB - In this paper, we propose a method for temporal segmentation of human repetitive actions based on frequency analysis of kinematic parameters, zero-velocity crossing detection, and adaptive k-means clustering. Since the human motion data may be captured with different modalities which have different temporal sampling rate and accuracy (e.g., Optical motion capture systems vs. Microsoft Kinect), we first apply a generic full-body kinematic model with an unscented Kalman filter to convert the motion data into a unified representation that is robust to noise. Furthermore, we extract the most representative kinematic parameters via the primary frequency analysis. The sequences are segmented based on zero-velocity crossing of the selected parameters followed by an adaptive k-means clustering to identify the repetition segments. Experimental results demonstrate that for the motion data captured by both the motion capture system and the Microsoft Kinect, our proposed algorithm obtains robust segmentation of repetitive action sequences.

KW - Bones

KW - Computer vision

KW - Extremities

KW - Hidden Markov models

KW - Joints

KW - Kinematics

KW - Motion segmentation

UR - http://www.scopus.com/inward/record.url?scp=84961751396&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84961751396&partnerID=8YFLogxK

U2 - 10.1109/3DV.2015.69

DO - 10.1109/3DV.2015.69

M3 - Conference contribution

SP - 562

EP - 570

BT - Proceedings - 2015 International Conference on 3D Vision, 3DV 2015

PB - Institute of Electrical and Electronics Engineers Inc.

ER -