Interpretable Gait Recognition by Granger Causality

D 2022

Interpretable Gait Recognition by Granger Causality

BALÁŽIA, Michal, Kateřina HLAVÁČKOVÁ-SCHINDLER, Petr SOJKA and Claudia PLANT

Basic information

Original name

Interpretable Gait Recognition by Granger Causality

Authors

BALÁŽIA, Michal (703 Slovakia, guarantor, belonging to the institution), Kateřina HLAVÁČKOVÁ-SCHINDLER (203 Czech Republic), Petr SOJKA (203 Czech Republic, belonging to the institution) and Claudia PLANT (40 Austria)

Edition

Los Alamitos, CA, USA, Proceedings of 26th International Conference on Pattern Recognition, ICPR 2022, p. 1069-1075, 7 pp. 2022

Publisher

IEEE

Other information

Language

English

Type of outcome

Stať ve sborníku

Field of Study

10201 Computer sciences, information science, bioinformatics

Country of publisher

Czech Republic

Confidentiality degree

není předmětem státního či obchodního tajemství

Publication form

printed version "print"

References:

preprint (arXiv) fulltext PDF (DOI)

RIV identification code

RIV/00216224:14330/22:00125672

Organization unit

Faculty of Informatics

ISBN

978-1-6654-9062-7

ISSN

DOI

http://dx.doi.org/10.1109/ICPR56361.2022.9956624

UT WoS

000897707601011

Keywords (in Czech)

Grangerova kauzalita; rozpoznávání podle chůze

Keywords in English

Granger causality; gait recognition

Abstract

V originále

Which joint interactions in the human gait cycle can be used as biometric characteristics? Most current methods on gait recognition suffer from a lack of interpretability. We propose an interpretable feature representation of gait sequences by the graphical Granger causal inference. The gait sequence of a person in the standardized motion capture format, constituting a set of 3D joint spatial trajectories, is envisaged as a causal system of joints interacting in time. We apply the graphical Granger model (GGM) to obtain the so-called Granger causal graph among joints as a discriminative and visually interpretable representation of a person's gait. We evaluate eleven distance functions in the GGM feature space by established classification and class-separability evaluation metrics. Our experiments indicate that, depending on the metric, the most appropriate distance functions for the GGM are the total norm distance and the Ky-Fan 1-norm distance. Experiments also show that the GGM is able to detect the most discriminative joint interactions and that it outperforms five related interpretable models in correct classification rate and in the Davies-Bouldin index. The proposed GGM model can serve as a complementary tool for gait analysis in kinesiology or for gait recognition in video surveillance.

Citovat

BALÁŽIA, Michal, Kateřina HLAVÁČKOVÁ-SCHINDLER, Petr SOJKA and Claudia PLANT. Interpretable Gait Recognition by Granger Causality. In Proceedings of 26th International Conference on Pattern Recognition, ICPR 2022. Los Alamitos, CA, USA: IEEE, 2022, p. 1069-1075. ISBN 978-1-6654-9062-7. Available from: https://dx.doi.org/10.1109/ICPR56361.2022.9956624.

@inproceedings{1847097,
   author = {Balážia, Michal and HlaváčkováandSchindler, Kateřina and Sojka, Petr and Plant, Claudia},
   address = {Los Alamitos, CA, USA},
   booktitle = {Proceedings of 26th International Conference on Pattern Recognition, ICPR 2022},
   doi = {http://dx.doi.org/10.1109/ICPR56361.2022.9956624},
   keywords = {Granger causality; gait recognition},
   howpublished = {tištěná verze "print"},
   language = {eng},
   location = {Los Alamitos, CA, USA},
   isbn = {978-1-6654-9062-7},
   pages = {1069-1075},
   publisher = {IEEE},
   title = {Interpretable Gait Recognition by Granger Causality},
   url = {https://arxiv.org/abs/2206.06714},
   year = {2022}
}

TY  - JOUR
ID  - 1847097
AU  - Balážia, Michal - Hlaváčková-Schindler, Kateřina - Sojka, Petr - Plant, Claudia
PY  - 2022
TI  - Interpretable Gait Recognition by Granger Causality
PB  - IEEE
CY  - Los Alamitos, CA, USA
SN  - 9781665490627
KW  - Granger causality
KW  - gait recognition
UR  - https://arxiv.org/abs/2206.06714
N2  - Which joint interactions in the human gait cycle can be used as biometric characteristics? Most current methods on gait recognition suffer from a lack of interpretability. We propose an interpretable feature representation of gait sequences by the graphical Granger causal inference. The gait sequence of a person in the standardized motion capture format, constituting a set of 3D joint spatial trajectories, is envisaged as a causal system of joints interacting in time. We apply the graphical Granger model (GGM) to obtain the so-called Granger causal graph among joints as a discriminative and visually interpretable representation of a person's gait. We evaluate eleven distance functions in the GGM feature space by established classification and class-separability evaluation metrics. Our experiments indicate that, depending on the metric, the most appropriate distance functions for the GGM are the total norm distance and the Ky-Fan 1-norm distance. Experiments also show that the GGM is able to detect the most discriminative joint interactions and that it outperforms five related interpretable models in correct classification rate and in the Davies-Bouldin index. The proposed GGM model can serve as a complementary tool for gait analysis in kinesiology or for gait recognition in video surveillance.
ER  -

BALÁŽIA, Michal, Kateřina HLAVÁČKOVÁ-SCHINDLER, Petr SOJKA and Claudia PLANT. Interpretable Gait Recognition by Granger Causality. In \textit{Proceedings of 26th International Conference on Pattern Recognition, ICPR 2022}. Los Alamitos, CA, USA: IEEE, 2022, p.~1069-1075. ISBN~978-1-6654-9062-7. Available from: https://dx.doi.org/10.1109/ICPR56361.2022.9956624.

Detailed Information on Publication Record