A single depth sensor based human activity recognition via convolutional neural network

J. H. Park; S. U. Park; Md Zia Uddin; M. A. Al-Antari; M. A. Al-Masni; T. S. Kim

doi:10.1007/978-981-10-4361-1_92

A single depth sensor based human activity recognition via convolutional neural network

J. H. Park
, S. U. Park
, Md Zia Uddin
, M. A. Al-Antari
, M. A. Al-Masni
, T. S. Kim^*

^*Corresponding author for this work

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution › peer-review

Abstract

Human activity recognition (HAR) has become an active research topic in the various fields. Depth sensor-based HAR recognizes human activities using features from depth human silhouettes via classifiers such as Hidden Markov Model (HMM), Conditional Random Fields Model etc. In this paper, we propose a new HAR system via Convolutional Neural Network (CNN), one of deep learning algorithms. We extract joint angles from multiple body joints changing in time and create a spatiotemporal feature matrix (i.e., multiple body joint angles in time). With these derived features, we train and test our CNN for HAR. In order to evaluate our system, we have compared the performance of our CNN-based HAR against the HMM- and Deep Belief Network (DBN)-based HAR using a database of Microsoft Research Cambridge-12 (MSRC-12). Our test results show that the proposed CNN-based HAR is able to recognize twelve human activities reliably and it outperforms the HMM- and DBN-based systems. We have achieved the average recognition accuracy of 98.59% for the activities. The results are 6.1% more accurate than that of the HMM-based HAR and 1.05% more accurate than that of the DBN-based HAR.

Original language	English
Title of host publication	6th International Conference on the Development of Biomedical Engineering in Vietnam, BME6
Editors	Toi Vo Van, Thanh An Nguyen Le, Thang Nguyen Duc
Publisher	Springer Verlag
Pages	541-545
Number of pages	5
ISBN (Print)	9789811043604
DOIs	https://doi.org/10.1007/978-981-10-4361-1_92
State	Published - 2018
Externally published	Yes
Event	6th International Conference on the Development of Biomedical Engineering in Vietnam, BME 2016 - Ho Chi Minh, Viet Nam Duration: 27 Jun 2016 → 29 Jun 2016

Publication series

Name	IFMBE Proceedings
Volume	63
ISSN (Print)	1680-0737

Conference

Conference	6th International Conference on the Development of Biomedical Engineering in Vietnam, BME 2016
Country/Territory	Viet Nam
City	Ho Chi Minh
Period	27/06/16 → 29/06/16

Bibliographical note

Publisher Copyright:
© Springer Nature Singapore Pte Ltd. 2018.

Keywords

Convolutional neural network
Deep learning
Depth imaging sensor
Human activity recognition

ASJC Scopus subject areas

Bioengineering
Biomedical Engineering

Access to Document

10.1007/978-981-10-4361-1_92

Cite this

Park, J. H., Park, S. U., Zia Uddin, M., Al-Antari, M. A., Al-Masni, M. A., & Kim, T. S. (2018). A single depth sensor based human activity recognition via convolutional neural network. In T. Vo Van, T. A. Nguyen Le, & T. Nguyen Duc (Eds.), 6th International Conference on the Development of Biomedical Engineering in Vietnam, BME6 (pp. 541-545). (IFMBE Proceedings; Vol. 63). Springer Verlag. https://doi.org/10.1007/978-981-10-4361-1_92

@inproceedings{aac8d1d1359143b5ad30f84727d88d3e,

title = "A single depth sensor based human activity recognition via convolutional neural network",

abstract = "Human activity recognition (HAR) has become an active research topic in the various fields. Depth sensor-based HAR recognizes human activities using features from depth human silhouettes via classifiers such as Hidden Markov Model (HMM), Conditional Random Fields Model etc. In this paper, we propose a new HAR system via Convolutional Neural Network (CNN), one of deep learning algorithms. We extract joint angles from multiple body joints changing in time and create a spatiotemporal feature matrix (i.e., multiple body joint angles in time). With these derived features, we train and test our CNN for HAR. In order to evaluate our system, we have compared the performance of our CNN-based HAR against the HMM- and Deep Belief Network (DBN)-based HAR using a database of Microsoft Research Cambridge-12 (MSRC-12). Our test results show that the proposed CNN-based HAR is able to recognize twelve human activities reliably and it outperforms the HMM- and DBN-based systems. We have achieved the average recognition accuracy of 98.59\% for the activities. The results are 6.1\% more accurate than that of the HMM-based HAR and 1.05\% more accurate than that of the DBN-based HAR.",

keywords = "Convolutional neural network, Deep learning, Depth imaging sensor, Human activity recognition",

author = "Park, \{J. H.\} and Park, \{S. U.\} and \{Zia Uddin\}, Md and Al-Antari, \{M. A.\} and Al-Masni, \{M. A.\} and Kim, \{T. S.\}",

note = "Publisher Copyright: {\textcopyright} Springer Nature Singapore Pte Ltd. 2018.; 6th International Conference on the Development of Biomedical Engineering in Vietnam, BME 2016 ; Conference date: 27-06-2016 Through 29-06-2016",

year = "2018",

doi = "10.1007/978-981-10-4361-1\_92",

language = "English",

isbn = "9789811043604",

series = "IFMBE Proceedings",

publisher = "Springer Verlag",

pages = "541--545",

editor = "\{Vo Van\}, Toi and \{Nguyen Le\}, \{Thanh An\} and \{Nguyen Duc\}, Thang",

booktitle = "6th International Conference on the Development of Biomedical Engineering in Vietnam, BME6",

address = "Germany",

}

Park, JH, Park, SU, Zia Uddin, M, Al-Antari, MA, Al-Masni, MA & Kim, TS 2018, A single depth sensor based human activity recognition via convolutional neural network. in T Vo Van, TA Nguyen Le & T Nguyen Duc (eds), 6th International Conference on the Development of Biomedical Engineering in Vietnam, BME6. IFMBE Proceedings, vol. 63, Springer Verlag, pp. 541-545, 6th International Conference on the Development of Biomedical Engineering in Vietnam, BME 2016, Ho Chi Minh, Viet Nam, 27/06/16. https://doi.org/10.1007/978-981-10-4361-1_92

A single depth sensor based human activity recognition via convolutional neural network. / Park, J. H.; Park, S. U.; Zia Uddin, Md et al.
6th International Conference on the Development of Biomedical Engineering in Vietnam, BME6. ed. / Toi Vo Van; Thanh An Nguyen Le; Thang Nguyen Duc. Springer Verlag, 2018. p. 541-545 (IFMBE Proceedings; Vol. 63).

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution › peer-review

TY - GEN

T1 - A single depth sensor based human activity recognition via convolutional neural network

AU - Park, J. H.

AU - Park, S. U.

AU - Zia Uddin, Md

AU - Al-Antari, M. A.

AU - Al-Masni, M. A.

AU - Kim, T. S.

N1 - Publisher Copyright: © Springer Nature Singapore Pte Ltd. 2018.

PY - 2018

Y1 - 2018

N2 - Human activity recognition (HAR) has become an active research topic in the various fields. Depth sensor-based HAR recognizes human activities using features from depth human silhouettes via classifiers such as Hidden Markov Model (HMM), Conditional Random Fields Model etc. In this paper, we propose a new HAR system via Convolutional Neural Network (CNN), one of deep learning algorithms. We extract joint angles from multiple body joints changing in time and create a spatiotemporal feature matrix (i.e., multiple body joint angles in time). With these derived features, we train and test our CNN for HAR. In order to evaluate our system, we have compared the performance of our CNN-based HAR against the HMM- and Deep Belief Network (DBN)-based HAR using a database of Microsoft Research Cambridge-12 (MSRC-12). Our test results show that the proposed CNN-based HAR is able to recognize twelve human activities reliably and it outperforms the HMM- and DBN-based systems. We have achieved the average recognition accuracy of 98.59% for the activities. The results are 6.1% more accurate than that of the HMM-based HAR and 1.05% more accurate than that of the DBN-based HAR.

AB - Human activity recognition (HAR) has become an active research topic in the various fields. Depth sensor-based HAR recognizes human activities using features from depth human silhouettes via classifiers such as Hidden Markov Model (HMM), Conditional Random Fields Model etc. In this paper, we propose a new HAR system via Convolutional Neural Network (CNN), one of deep learning algorithms. We extract joint angles from multiple body joints changing in time and create a spatiotemporal feature matrix (i.e., multiple body joint angles in time). With these derived features, we train and test our CNN for HAR. In order to evaluate our system, we have compared the performance of our CNN-based HAR against the HMM- and Deep Belief Network (DBN)-based HAR using a database of Microsoft Research Cambridge-12 (MSRC-12). Our test results show that the proposed CNN-based HAR is able to recognize twelve human activities reliably and it outperforms the HMM- and DBN-based systems. We have achieved the average recognition accuracy of 98.59% for the activities. The results are 6.1% more accurate than that of the HMM-based HAR and 1.05% more accurate than that of the DBN-based HAR.

KW - Convolutional neural network

KW - Deep learning

KW - Depth imaging sensor

KW - Human activity recognition

UR - https://www.scopus.com/pages/publications/85030873681

U2 - 10.1007/978-981-10-4361-1_92

DO - 10.1007/978-981-10-4361-1_92

M3 - Conference contribution

AN - SCOPUS:85030873681

SN - 9789811043604

T3 - IFMBE Proceedings

SP - 541

EP - 545

BT - 6th International Conference on the Development of Biomedical Engineering in Vietnam, BME6

A2 - Vo Van, Toi

A2 - Nguyen Le, Thanh An

A2 - Nguyen Duc, Thang

PB - Springer Verlag

T2 - 6th International Conference on the Development of Biomedical Engineering in Vietnam, BME 2016

Y2 - 27 June 2016 through 29 June 2016

ER -

Park JH, Park SU, Zia Uddin M, Al-Antari MA, Al-Masni MA, Kim TS. A single depth sensor based human activity recognition via convolutional neural network. In Vo Van T, Nguyen Le TA, Nguyen Duc T, editors, 6th International Conference on the Development of Biomedical Engineering in Vietnam, BME6. Springer Verlag. 2018. p. 541-545. (IFMBE Proceedings). doi: 10.1007/978-981-10-4361-1_92

A single depth sensor based human activity recognition via convolutional neural network

Abstract

Publication series

Conference

Bibliographical note

Keywords

ASJC Scopus subject areas

Access to Document

Fingerprint

Cite this