Mode-Driven Volume Analysis Based on Correlation of Time Series

Chengcheng Jia, Wei Pang, Yun Fu

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

1 Citation (Scopus)

Abstract

Tensor analysis is widely used for face recognition and action recognition. In this paper, a mode-driven discriminant analysis (MDA) in tensor subspace is proposed for visual recognition. For training, we treat each sample as an N-order tensor, of which the first N-1 modes capture the spatial information of images while the N-th mode captures the sequential patterns of images. We employ Fisher criteria on the first N-1 modes to extract discriminative features of the visual information. After that, considering the correlation of adjacent frames in the sequence, i.e., the current frame and its former and latter ones, we update the sequence by calculating the correlation of triple adjacent frames, then perform discriminant analysis on the N-th mode. The alternating projection procedure of MDA converges and is convex with different initial values of the transformation matrices. Such hybrid tensor subspace learning scheme may sufficiently preserve both discrete and continuous distributions information of action videos in lower dimensional spaces to boost discriminant power. Experiments on the MSR action 3D database, KTH database and ETH database showed that our algorithm MDA outperformed other tensor-based methods in terms of accuracy and is competitive considering the time efficiency. Besides, it is robust to deal with the damaged and self-occluded action silhouettes and RGB object images in various viewing angles.
Original languageEnglish
Title of host publicationComputer Vision - ECCV 2014 Workshops
Subtitle of host publicationZurich, Switzerland, September 6-7 and 12, 2014, Proceedings, Part I
EditorsLourdes Agapito, Michael M. Bronstein, Carsten Rother
Place of PublicationZurich
PublisherSpringer
Pages818-833
Number of pages16
ISBN (Electronic)978-3-319-16178-5
ISBN (Print)978-3-319-16177-8
DOIs
Publication statusPublished - 19 Mar 2015

Publication series

NameLecture Notes in Computing Science
PublisherSpringer
Volume8925
ISSN (Print)0302-9743

Fingerprint

Tensors
Time series
Discriminant analysis
Face recognition
Experiments

Keywords

  • action recognition
  • canonical correlation analysis
  • adjacent frame
  • human action recognition
  • gait recognition

Cite this

Jia, C., Pang, W., & Fu, Y. (2015). Mode-Driven Volume Analysis Based on Correlation of Time Series. In L. Agapito, M. M. Bronstein, & C. Rother (Eds.), Computer Vision - ECCV 2014 Workshops: Zurich, Switzerland, September 6-7 and 12, 2014, Proceedings, Part I (pp. 818-833). (Lecture Notes in Computing Science; Vol. 8925). Zurich: Springer . https://doi.org/10.1007/978-3-319-16178-5_57

Mode-Driven Volume Analysis Based on Correlation of Time Series. / Jia, Chengcheng; Pang, Wei; Fu, Yun.

Computer Vision - ECCV 2014 Workshops: Zurich, Switzerland, September 6-7 and 12, 2014, Proceedings, Part I. ed. / Lourdes Agapito; Michael M. Bronstein; Carsten Rother. Zurich : Springer , 2015. p. 818-833 (Lecture Notes in Computing Science; Vol. 8925).

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

Jia, C, Pang, W & Fu, Y 2015, Mode-Driven Volume Analysis Based on Correlation of Time Series. in L Agapito, MM Bronstein & C Rother (eds), Computer Vision - ECCV 2014 Workshops: Zurich, Switzerland, September 6-7 and 12, 2014, Proceedings, Part I. Lecture Notes in Computing Science, vol. 8925, Springer , Zurich, pp. 818-833. https://doi.org/10.1007/978-3-319-16178-5_57
Jia C, Pang W, Fu Y. Mode-Driven Volume Analysis Based on Correlation of Time Series. In Agapito L, Bronstein MM, Rother C, editors, Computer Vision - ECCV 2014 Workshops: Zurich, Switzerland, September 6-7 and 12, 2014, Proceedings, Part I. Zurich: Springer . 2015. p. 818-833. (Lecture Notes in Computing Science). https://doi.org/10.1007/978-3-319-16178-5_57
Jia, Chengcheng ; Pang, Wei ; Fu, Yun. / Mode-Driven Volume Analysis Based on Correlation of Time Series. Computer Vision - ECCV 2014 Workshops: Zurich, Switzerland, September 6-7 and 12, 2014, Proceedings, Part I. editor / Lourdes Agapito ; Michael M. Bronstein ; Carsten Rother. Zurich : Springer , 2015. pp. 818-833 (Lecture Notes in Computing Science).
@inproceedings{6a446c1645d443dba3b454b6451901ca,
title = "Mode-Driven Volume Analysis Based on Correlation of Time Series",
abstract = "Tensor analysis is widely used for face recognition and action recognition. In this paper, a mode-driven discriminant analysis (MDA) in tensor subspace is proposed for visual recognition. For training, we treat each sample as an N-order tensor, of which the first N-1 modes capture the spatial information of images while the N-th mode captures the sequential patterns of images. We employ Fisher criteria on the first N-1 modes to extract discriminative features of the visual information. After that, considering the correlation of adjacent frames in the sequence, i.e., the current frame and its former and latter ones, we update the sequence by calculating the correlation of triple adjacent frames, then perform discriminant analysis on the N-th mode. The alternating projection procedure of MDA converges and is convex with different initial values of the transformation matrices. Such hybrid tensor subspace learning scheme may sufficiently preserve both discrete and continuous distributions information of action videos in lower dimensional spaces to boost discriminant power. Experiments on the MSR action 3D database, KTH database and ETH database showed that our algorithm MDA outperformed other tensor-based methods in terms of accuracy and is competitive considering the time efficiency. Besides, it is robust to deal with the damaged and self-occluded action silhouettes and RGB object images in various viewing angles.",
keywords = "action recognition, canonical correlation analysis, adjacent frame, human action recognition, gait recognition",
author = "Chengcheng Jia and Wei Pang and Yun Fu",
year = "2015",
month = "3",
day = "19",
doi = "10.1007/978-3-319-16178-5_57",
language = "English",
isbn = "978-3-319-16177-8",
series = "Lecture Notes in Computing Science",
publisher = "Springer",
pages = "818--833",
editor = "Lourdes Agapito and Bronstein, {Michael M.} and Carsten Rother",
booktitle = "Computer Vision - ECCV 2014 Workshops",

}

TY - GEN

T1 - Mode-Driven Volume Analysis Based on Correlation of Time Series

AU - Jia, Chengcheng

AU - Pang, Wei

AU - Fu, Yun

PY - 2015/3/19

Y1 - 2015/3/19

N2 - Tensor analysis is widely used for face recognition and action recognition. In this paper, a mode-driven discriminant analysis (MDA) in tensor subspace is proposed for visual recognition. For training, we treat each sample as an N-order tensor, of which the first N-1 modes capture the spatial information of images while the N-th mode captures the sequential patterns of images. We employ Fisher criteria on the first N-1 modes to extract discriminative features of the visual information. After that, considering the correlation of adjacent frames in the sequence, i.e., the current frame and its former and latter ones, we update the sequence by calculating the correlation of triple adjacent frames, then perform discriminant analysis on the N-th mode. The alternating projection procedure of MDA converges and is convex with different initial values of the transformation matrices. Such hybrid tensor subspace learning scheme may sufficiently preserve both discrete and continuous distributions information of action videos in lower dimensional spaces to boost discriminant power. Experiments on the MSR action 3D database, KTH database and ETH database showed that our algorithm MDA outperformed other tensor-based methods in terms of accuracy and is competitive considering the time efficiency. Besides, it is robust to deal with the damaged and self-occluded action silhouettes and RGB object images in various viewing angles.

AB - Tensor analysis is widely used for face recognition and action recognition. In this paper, a mode-driven discriminant analysis (MDA) in tensor subspace is proposed for visual recognition. For training, we treat each sample as an N-order tensor, of which the first N-1 modes capture the spatial information of images while the N-th mode captures the sequential patterns of images. We employ Fisher criteria on the first N-1 modes to extract discriminative features of the visual information. After that, considering the correlation of adjacent frames in the sequence, i.e., the current frame and its former and latter ones, we update the sequence by calculating the correlation of triple adjacent frames, then perform discriminant analysis on the N-th mode. The alternating projection procedure of MDA converges and is convex with different initial values of the transformation matrices. Such hybrid tensor subspace learning scheme may sufficiently preserve both discrete and continuous distributions information of action videos in lower dimensional spaces to boost discriminant power. Experiments on the MSR action 3D database, KTH database and ETH database showed that our algorithm MDA outperformed other tensor-based methods in terms of accuracy and is competitive considering the time efficiency. Besides, it is robust to deal with the damaged and self-occluded action silhouettes and RGB object images in various viewing angles.

KW - action recognition

KW - canonical correlation analysis

KW - adjacent frame

KW - human action recognition

KW - gait recognition

U2 - 10.1007/978-3-319-16178-5_57

DO - 10.1007/978-3-319-16178-5_57

M3 - Conference contribution

SN - 978-3-319-16177-8

T3 - Lecture Notes in Computing Science

SP - 818

EP - 833

BT - Computer Vision - ECCV 2014 Workshops

A2 - Agapito, Lourdes

A2 - Bronstein, Michael M.

A2 - Rother, Carsten

PB - Springer

CY - Zurich

ER -