Accelerometer dense trajectories for activity recognition and people identification

Roberto Leyva, Geise Santos, Anderson Rocha, Victor Sanchez, Chang Tsun Li

Research output: Book chapter/Published conference paperConference paper

Abstract

This paper addresses the problem of activity recognition and people identification using accelerometer signals acquired by personal devices. Specifically, we propose a framework based on a Deep Neural Network that employs an efficient dense trajectory encoding to compute features. These Accelerometer Dense Trajectory (ADT) features, which are similar to those used for action recognition in the spatio-Temporal domain of video data, densely map the accelerometer signals into three-dimensional normalised positions. To deal with the unordered nature and dimensional variation of trajectories associated with the classes, the proposed framework employs Fisher Vectors as a high order representation of the extracted features. We evaluate the proposed ADT features and framework on the Sphere2016 Challenge and WISDM datasets for activity recognition. For people identification, we employ the RecodGait dataset. For these two significantly different classification tasks, the performance evaluation results confirm the high descriptiveness of the proposed ADT features and the effectiveness of the proposed framework.

Original languageEnglish
Title of host publication2019 7th International Workshop on Biometrics and Forensics, IWBF 2019
PublisherIEEE, Institute of Electrical and Electronics Engineers
Number of pages6
ISBN (Electronic)9781728106229
ISBN (Print)9781728106236
DOIs
Publication statusPublished - Jun 2019
Event7th International Workshop on Biometrics and Forensics: IWBF 2019 - Centro de Educacion Continua, Cancun, Mexico
Duration: 02 May 201903 May 2019
https://warwick.ac.uk/fac/sci/dcs/people/victor_sanchez/iwbf2019/ (conference website)

Publication series

Name2019 7th International Workshop on Biometrics and Forensics, IWBF 2019

Conference

Conference7th International Workshop on Biometrics and Forensics
CountryMexico
CityCancun
Period02/05/1903/05/19
Internet address

Fingerprint

accelerometers
Accelerometers
Trajectories
trajectories
Task Performance and Analysis
Equipment and Supplies
video data
coding
Datasets
evaluation

Cite this

Leyva, R., Santos, G., Rocha, A., Sanchez, V., & Li, C. T. (2019). Accelerometer dense trajectories for activity recognition and people identification. In 2019 7th International Workshop on Biometrics and Forensics, IWBF 2019 [8739218] (2019 7th International Workshop on Biometrics and Forensics, IWBF 2019). IEEE, Institute of Electrical and Electronics Engineers. https://doi.org/10.1109/IWBF.2019.8739218
Leyva, Roberto ; Santos, Geise ; Rocha, Anderson ; Sanchez, Victor ; Li, Chang Tsun. / Accelerometer dense trajectories for activity recognition and people identification. 2019 7th International Workshop on Biometrics and Forensics, IWBF 2019. IEEE, Institute of Electrical and Electronics Engineers, 2019. (2019 7th International Workshop on Biometrics and Forensics, IWBF 2019).
@inproceedings{3bc4e597f33d42de86e344c2660f44b3,
title = "Accelerometer dense trajectories for activity recognition and people identification",
abstract = "This paper addresses the problem of activity recognition and people identification using accelerometer signals acquired by personal devices. Specifically, we propose a framework based on a Deep Neural Network that employs an efficient dense trajectory encoding to compute features. These Accelerometer Dense Trajectory (ADT) features, which are similar to those used for action recognition in the spatio-Temporal domain of video data, densely map the accelerometer signals into three-dimensional normalised positions. To deal with the unordered nature and dimensional variation of trajectories associated with the classes, the proposed framework employs Fisher Vectors as a high order representation of the extracted features. We evaluate the proposed ADT features and framework on the Sphere2016 Challenge and WISDM datasets for activity recognition. For people identification, we employ the RecodGait dataset. For these two significantly different classification tasks, the performance evaluation results confirm the high descriptiveness of the proposed ADT features and the effectiveness of the proposed framework.",
keywords = "accelerometer sensors, Deep Neural Network (DNN), Dense trajectories",
author = "Roberto Leyva and Geise Santos and Anderson Rocha and Victor Sanchez and Li, {Chang Tsun}",
year = "2019",
month = "6",
doi = "10.1109/IWBF.2019.8739218",
language = "English",
isbn = "9781728106236",
series = "2019 7th International Workshop on Biometrics and Forensics, IWBF 2019",
publisher = "IEEE, Institute of Electrical and Electronics Engineers",
booktitle = "2019 7th International Workshop on Biometrics and Forensics, IWBF 2019",
address = "United States",

}

Leyva, R, Santos, G, Rocha, A, Sanchez, V & Li, CT 2019, Accelerometer dense trajectories for activity recognition and people identification. in 2019 7th International Workshop on Biometrics and Forensics, IWBF 2019., 8739218, 2019 7th International Workshop on Biometrics and Forensics, IWBF 2019, IEEE, Institute of Electrical and Electronics Engineers, 7th International Workshop on Biometrics and Forensics, Cancun, Mexico, 02/05/19. https://doi.org/10.1109/IWBF.2019.8739218

Accelerometer dense trajectories for activity recognition and people identification. / Leyva, Roberto; Santos, Geise; Rocha, Anderson; Sanchez, Victor; Li, Chang Tsun.

2019 7th International Workshop on Biometrics and Forensics, IWBF 2019. IEEE, Institute of Electrical and Electronics Engineers, 2019. 8739218 (2019 7th International Workshop on Biometrics and Forensics, IWBF 2019).

Research output: Book chapter/Published conference paperConference paper

TY - GEN

T1 - Accelerometer dense trajectories for activity recognition and people identification

AU - Leyva, Roberto

AU - Santos, Geise

AU - Rocha, Anderson

AU - Sanchez, Victor

AU - Li, Chang Tsun

PY - 2019/6

Y1 - 2019/6

N2 - This paper addresses the problem of activity recognition and people identification using accelerometer signals acquired by personal devices. Specifically, we propose a framework based on a Deep Neural Network that employs an efficient dense trajectory encoding to compute features. These Accelerometer Dense Trajectory (ADT) features, which are similar to those used for action recognition in the spatio-Temporal domain of video data, densely map the accelerometer signals into three-dimensional normalised positions. To deal with the unordered nature and dimensional variation of trajectories associated with the classes, the proposed framework employs Fisher Vectors as a high order representation of the extracted features. We evaluate the proposed ADT features and framework on the Sphere2016 Challenge and WISDM datasets for activity recognition. For people identification, we employ the RecodGait dataset. For these two significantly different classification tasks, the performance evaluation results confirm the high descriptiveness of the proposed ADT features and the effectiveness of the proposed framework.

AB - This paper addresses the problem of activity recognition and people identification using accelerometer signals acquired by personal devices. Specifically, we propose a framework based on a Deep Neural Network that employs an efficient dense trajectory encoding to compute features. These Accelerometer Dense Trajectory (ADT) features, which are similar to those used for action recognition in the spatio-Temporal domain of video data, densely map the accelerometer signals into three-dimensional normalised positions. To deal with the unordered nature and dimensional variation of trajectories associated with the classes, the proposed framework employs Fisher Vectors as a high order representation of the extracted features. We evaluate the proposed ADT features and framework on the Sphere2016 Challenge and WISDM datasets for activity recognition. For people identification, we employ the RecodGait dataset. For these two significantly different classification tasks, the performance evaluation results confirm the high descriptiveness of the proposed ADT features and the effectiveness of the proposed framework.

KW - accelerometer sensors

KW - Deep Neural Network (DNN)

KW - Dense trajectories

UR - http://www.scopus.com/inward/record.url?scp=85068481497&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85068481497&partnerID=8YFLogxK

U2 - 10.1109/IWBF.2019.8739218

DO - 10.1109/IWBF.2019.8739218

M3 - Conference paper

AN - SCOPUS:85068481497

SN - 9781728106236

T3 - 2019 7th International Workshop on Biometrics and Forensics, IWBF 2019

BT - 2019 7th International Workshop on Biometrics and Forensics, IWBF 2019

PB - IEEE, Institute of Electrical and Electronics Engineers

ER -

Leyva R, Santos G, Rocha A, Sanchez V, Li CT. Accelerometer dense trajectories for activity recognition and people identification. In 2019 7th International Workshop on Biometrics and Forensics, IWBF 2019. IEEE, Institute of Electrical and Electronics Engineers. 2019. 8739218. (2019 7th International Workshop on Biometrics and Forensics, IWBF 2019). https://doi.org/10.1109/IWBF.2019.8739218