Deep convolutional neural networks for human activity recognition with smartphone sensors

Charissa Ann Ronao, Sung Bae Cho

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

45 Citations (Scopus)

Abstract

Human activity recognition (HAR) using smartphone sensors utilize time-series, multivariate data to detect activities. Time-series data have inherent local dependency characteristics. Moreover, activities tend to be hierarchical and translation invariant in nature. Consequently, convolutional neural networks (convnet) exploit these characteristics, which make it appropriate in dealing with time-series sensor data. In this paper, we propose an architecture of convnets with sensor data gathered from smartphone sensors to recognize activities. Experiments show that increasing the number of convolutional layers increases performance, but the complexity of the derived features decreases with every additional layer. Moreover, preserving the information passed from layer to layer is more important, as opposed to blindly increasing the hyperparameters to improve performance. The convnet structure can also benefit from a wider filter size and lower pooling size setting. Lastly, we show that convnet outperforms all the other state-of-the-art techniques in HAR, especially SVM, which achieved the previous best result for the data set.

Original languageEnglish
Title of host publicationNeural Information Processing - 22nd International Conference, ICONIP 2015, Proceedings
EditorsSabri Arik, Tingwen Huang, Weng Kin Lai, Qingshan Liu
PublisherSpringer Verlag
Pages46-53
Number of pages8
ISBN (Print)9783319265605
DOIs
Publication statusPublished - 2015 Jan 1
Event22nd International Conference on Neural Information Processing, ICONIP 2015 - Istanbul, Turkey
Duration: 2015 Nov 92015 Nov 12

Publication series

NameLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume9492
ISSN (Print)0302-9743
ISSN (Electronic)1611-3349

Other

Other22nd International Conference on Neural Information Processing, ICONIP 2015
CountryTurkey
CityIstanbul
Period15/11/915/11/12

Fingerprint

Activity Recognition
Smartphones
Neural Networks
Neural networks
Time series
Sensor
Sensors
Hyperparameters
Pooling
Multivariate Data
Time Series Data
Network Structure
Tend
Filter
Decrease
Invariant
Human
Experiment
Experiments

All Science Journal Classification (ASJC) codes

  • Theoretical Computer Science
  • Computer Science(all)

Cite this

Ronao, C. A., & Cho, S. B. (2015). Deep convolutional neural networks for human activity recognition with smartphone sensors. In S. Arik, T. Huang, W. K. Lai, & Q. Liu (Eds.), Neural Information Processing - 22nd International Conference, ICONIP 2015, Proceedings (pp. 46-53). (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 9492). Springer Verlag. https://doi.org/10.1007/978-3-319-26561-2_6
Ronao, Charissa Ann ; Cho, Sung Bae. / Deep convolutional neural networks for human activity recognition with smartphone sensors. Neural Information Processing - 22nd International Conference, ICONIP 2015, Proceedings. editor / Sabri Arik ; Tingwen Huang ; Weng Kin Lai ; Qingshan Liu. Springer Verlag, 2015. pp. 46-53 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)).
@inproceedings{0370b61510644542a0aecd2ab0797b0a,
title = "Deep convolutional neural networks for human activity recognition with smartphone sensors",
abstract = "Human activity recognition (HAR) using smartphone sensors utilize time-series, multivariate data to detect activities. Time-series data have inherent local dependency characteristics. Moreover, activities tend to be hierarchical and translation invariant in nature. Consequently, convolutional neural networks (convnet) exploit these characteristics, which make it appropriate in dealing with time-series sensor data. In this paper, we propose an architecture of convnets with sensor data gathered from smartphone sensors to recognize activities. Experiments show that increasing the number of convolutional layers increases performance, but the complexity of the derived features decreases with every additional layer. Moreover, preserving the information passed from layer to layer is more important, as opposed to blindly increasing the hyperparameters to improve performance. The convnet structure can also benefit from a wider filter size and lower pooling size setting. Lastly, we show that convnet outperforms all the other state-of-the-art techniques in HAR, especially SVM, which achieved the previous best result for the data set.",
author = "Ronao, {Charissa Ann} and Cho, {Sung Bae}",
year = "2015",
month = "1",
day = "1",
doi = "10.1007/978-3-319-26561-2_6",
language = "English",
isbn = "9783319265605",
series = "Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)",
publisher = "Springer Verlag",
pages = "46--53",
editor = "Sabri Arik and Tingwen Huang and Lai, {Weng Kin} and Qingshan Liu",
booktitle = "Neural Information Processing - 22nd International Conference, ICONIP 2015, Proceedings",
address = "Germany",

}

Ronao, CA & Cho, SB 2015, Deep convolutional neural networks for human activity recognition with smartphone sensors. in S Arik, T Huang, WK Lai & Q Liu (eds), Neural Information Processing - 22nd International Conference, ICONIP 2015, Proceedings. Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), vol. 9492, Springer Verlag, pp. 46-53, 22nd International Conference on Neural Information Processing, ICONIP 2015, Istanbul, Turkey, 15/11/9. https://doi.org/10.1007/978-3-319-26561-2_6

Deep convolutional neural networks for human activity recognition with smartphone sensors. / Ronao, Charissa Ann; Cho, Sung Bae.

Neural Information Processing - 22nd International Conference, ICONIP 2015, Proceedings. ed. / Sabri Arik; Tingwen Huang; Weng Kin Lai; Qingshan Liu. Springer Verlag, 2015. p. 46-53 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 9492).

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

TY - GEN

T1 - Deep convolutional neural networks for human activity recognition with smartphone sensors

AU - Ronao, Charissa Ann

AU - Cho, Sung Bae

PY - 2015/1/1

Y1 - 2015/1/1

N2 - Human activity recognition (HAR) using smartphone sensors utilize time-series, multivariate data to detect activities. Time-series data have inherent local dependency characteristics. Moreover, activities tend to be hierarchical and translation invariant in nature. Consequently, convolutional neural networks (convnet) exploit these characteristics, which make it appropriate in dealing with time-series sensor data. In this paper, we propose an architecture of convnets with sensor data gathered from smartphone sensors to recognize activities. Experiments show that increasing the number of convolutional layers increases performance, but the complexity of the derived features decreases with every additional layer. Moreover, preserving the information passed from layer to layer is more important, as opposed to blindly increasing the hyperparameters to improve performance. The convnet structure can also benefit from a wider filter size and lower pooling size setting. Lastly, we show that convnet outperforms all the other state-of-the-art techniques in HAR, especially SVM, which achieved the previous best result for the data set.

AB - Human activity recognition (HAR) using smartphone sensors utilize time-series, multivariate data to detect activities. Time-series data have inherent local dependency characteristics. Moreover, activities tend to be hierarchical and translation invariant in nature. Consequently, convolutional neural networks (convnet) exploit these characteristics, which make it appropriate in dealing with time-series sensor data. In this paper, we propose an architecture of convnets with sensor data gathered from smartphone sensors to recognize activities. Experiments show that increasing the number of convolutional layers increases performance, but the complexity of the derived features decreases with every additional layer. Moreover, preserving the information passed from layer to layer is more important, as opposed to blindly increasing the hyperparameters to improve performance. The convnet structure can also benefit from a wider filter size and lower pooling size setting. Lastly, we show that convnet outperforms all the other state-of-the-art techniques in HAR, especially SVM, which achieved the previous best result for the data set.

UR - http://www.scopus.com/inward/record.url?scp=84951871458&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84951871458&partnerID=8YFLogxK

U2 - 10.1007/978-3-319-26561-2_6

DO - 10.1007/978-3-319-26561-2_6

M3 - Conference contribution

AN - SCOPUS:84951871458

SN - 9783319265605

T3 - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)

SP - 46

EP - 53

BT - Neural Information Processing - 22nd International Conference, ICONIP 2015, Proceedings

A2 - Arik, Sabri

A2 - Huang, Tingwen

A2 - Lai, Weng Kin

A2 - Liu, Qingshan

PB - Springer Verlag

ER -

Ronao CA, Cho SB. Deep convolutional neural networks for human activity recognition with smartphone sensors. In Arik S, Huang T, Lai WK, Liu Q, editors, Neural Information Processing - 22nd International Conference, ICONIP 2015, Proceedings. Springer Verlag. 2015. p. 46-53. (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)). https://doi.org/10.1007/978-3-319-26561-2_6