A Public Domain Dataset for Human Activity Recognition in Free-living Conditions

Federico Cruciani, Chen Sun, Shuai Zhang, CD Nugent, Chunping Li, Shaoxu Song, Cheng Cheng, I Cleland, P McCullagh

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

Abstract

In Human Activity Recognition (HAR), supervised
Machine Learning methods are predominantly used, making
availability of datasets a major issue for research in the field.
In particular, the majority of available datasets are collected
under controlled conditions. Consequently, models trained under
similar circumstances, generally exhibit a significant decrease in
recognition accuracy when they are moved to final deployment
in the wild, within unconstrained settings. This paper presents a
new dataset for HAR, collected in free-living and unconstrained
conditions. 10 subjects were recruited for the purpose of data
collection. Data was recorded over a 6 week period using
a smartphone app, and a wristband activity monitor. During
the first and last week of observation, participants also wore
an ActivPAL™ activity logger. The data collected have been
partially self-labeled by participants, by means of the mobile app
provided for data collection. The dataset collected can be used to
evaluate HAR algorithm and models in real-world unconstrained
settings. Together with the description of the dataset, this work
presents some preliminary results, obtained cross-validating a
model trained on the publicly available Extrasensory dataset, and
testing its performance on our newly collected dataset. Results
obtained highlighted high cross-subject variability when testing
on new subjects, with a balanced accuracy varying between
53.33% and 90.01%, with an average balanced accuracy of
71.73%.
LanguageEnglish
Title of host publicationThe 2nd Workshop on advanced Technologies for Smarter Assisted Living solutions: Towards an open Smart Home infrastructure (SmarterAAL'19)
Pages166-171
Number of pages6
Publication statusAccepted/In press - 21 May 2019

Fingerprint

Smartphones
Application programs

Keywords

  • Balanced Batch Learning
  • Convolutional Neural Networks
  • Dataset
  • Free-living
  • Human Activity Recognition
  • Machine Learning

Cite this

Cruciani, F., Sun, C., Zhang, S., Nugent, CD., Li, C., Song, S., ... McCullagh, P. (Accepted/In press). A Public Domain Dataset for Human Activity Recognition in Free-living Conditions. In The 2nd Workshop on advanced Technologies for Smarter Assisted Living solutions: Towards an open Smart Home infrastructure (SmarterAAL'19) (pp. 166-171)
Cruciani, Federico ; Sun, Chen ; Zhang, Shuai ; Nugent, CD ; Li, Chunping ; Song, Shaoxu ; Cheng, Cheng ; Cleland, I ; McCullagh, P. / A Public Domain Dataset for Human Activity Recognition in Free-living Conditions. The 2nd Workshop on advanced Technologies for Smarter Assisted Living solutions: Towards an open Smart Home infrastructure (SmarterAAL'19). 2019. pp. 166-171
@inproceedings{9d9cb5faaecf47fa9b4d80ea461b1950,
title = "A Public Domain Dataset for Human Activity Recognition in Free-living Conditions",
abstract = "In Human Activity Recognition (HAR), supervisedMachine Learning methods are predominantly used, makingavailability of datasets a major issue for research in the field.In particular, the majority of available datasets are collectedunder controlled conditions. Consequently, models trained undersimilar circumstances, generally exhibit a significant decrease inrecognition accuracy when they are moved to final deploymentin the wild, within unconstrained settings. This paper presents anew dataset for HAR, collected in free-living and unconstrainedconditions. 10 subjects were recruited for the purpose of datacollection. Data was recorded over a 6 week period usinga smartphone app, and a wristband activity monitor. Duringthe first and last week of observation, participants also worean ActivPAL™ activity logger. The data collected have beenpartially self-labeled by participants, by means of the mobile appprovided for data collection. The dataset collected can be used toevaluate HAR algorithm and models in real-world unconstrainedsettings. Together with the description of the dataset, this workpresents some preliminary results, obtained cross-validating amodel trained on the publicly available Extrasensory dataset, andtesting its performance on our newly collected dataset. Resultsobtained highlighted high cross-subject variability when testingon new subjects, with a balanced accuracy varying between53.33{\%} and 90.01{\%}, with an average balanced accuracy of71.73{\%}.",
keywords = "Balanced Batch Learning, Convolutional Neural Networks, Dataset, Free-living, Human Activity Recognition, Machine Learning",
author = "Federico Cruciani and Chen Sun and Shuai Zhang and CD Nugent and Chunping Li and Shaoxu Song and Cheng Cheng and I Cleland and P McCullagh",
year = "2019",
month = "5",
day = "21",
language = "English",
pages = "166--171",
booktitle = "The 2nd Workshop on advanced Technologies for Smarter Assisted Living solutions: Towards an open Smart Home infrastructure (SmarterAAL'19)",

}

Cruciani, F, Sun, C, Zhang, S, Nugent, CD, Li, C, Song, S, Cheng, C, Cleland, I & McCullagh, P 2019, A Public Domain Dataset for Human Activity Recognition in Free-living Conditions. in The 2nd Workshop on advanced Technologies for Smarter Assisted Living solutions: Towards an open Smart Home infrastructure (SmarterAAL'19). pp. 166-171.

A Public Domain Dataset for Human Activity Recognition in Free-living Conditions. / Cruciani, Federico; Sun, Chen; Zhang, Shuai; Nugent, CD; Li, Chunping; Song, Shaoxu; Cheng, Cheng; Cleland, I; McCullagh, P.

The 2nd Workshop on advanced Technologies for Smarter Assisted Living solutions: Towards an open Smart Home infrastructure (SmarterAAL'19). 2019. p. 166-171.

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

TY - GEN

T1 - A Public Domain Dataset for Human Activity Recognition in Free-living Conditions

AU - Cruciani, Federico

AU - Sun, Chen

AU - Zhang, Shuai

AU - Nugent, CD

AU - Li, Chunping

AU - Song, Shaoxu

AU - Cheng, Cheng

AU - Cleland, I

AU - McCullagh, P

PY - 2019/5/21

Y1 - 2019/5/21

N2 - In Human Activity Recognition (HAR), supervisedMachine Learning methods are predominantly used, makingavailability of datasets a major issue for research in the field.In particular, the majority of available datasets are collectedunder controlled conditions. Consequently, models trained undersimilar circumstances, generally exhibit a significant decrease inrecognition accuracy when they are moved to final deploymentin the wild, within unconstrained settings. This paper presents anew dataset for HAR, collected in free-living and unconstrainedconditions. 10 subjects were recruited for the purpose of datacollection. Data was recorded over a 6 week period usinga smartphone app, and a wristband activity monitor. Duringthe first and last week of observation, participants also worean ActivPAL™ activity logger. The data collected have beenpartially self-labeled by participants, by means of the mobile appprovided for data collection. The dataset collected can be used toevaluate HAR algorithm and models in real-world unconstrainedsettings. Together with the description of the dataset, this workpresents some preliminary results, obtained cross-validating amodel trained on the publicly available Extrasensory dataset, andtesting its performance on our newly collected dataset. Resultsobtained highlighted high cross-subject variability when testingon new subjects, with a balanced accuracy varying between53.33% and 90.01%, with an average balanced accuracy of71.73%.

AB - In Human Activity Recognition (HAR), supervisedMachine Learning methods are predominantly used, makingavailability of datasets a major issue for research in the field.In particular, the majority of available datasets are collectedunder controlled conditions. Consequently, models trained undersimilar circumstances, generally exhibit a significant decrease inrecognition accuracy when they are moved to final deploymentin the wild, within unconstrained settings. This paper presents anew dataset for HAR, collected in free-living and unconstrainedconditions. 10 subjects were recruited for the purpose of datacollection. Data was recorded over a 6 week period usinga smartphone app, and a wristband activity monitor. Duringthe first and last week of observation, participants also worean ActivPAL™ activity logger. The data collected have beenpartially self-labeled by participants, by means of the mobile appprovided for data collection. The dataset collected can be used toevaluate HAR algorithm and models in real-world unconstrainedsettings. Together with the description of the dataset, this workpresents some preliminary results, obtained cross-validating amodel trained on the publicly available Extrasensory dataset, andtesting its performance on our newly collected dataset. Resultsobtained highlighted high cross-subject variability when testingon new subjects, with a balanced accuracy varying between53.33% and 90.01%, with an average balanced accuracy of71.73%.

KW - Balanced Batch Learning

KW - Convolutional Neural Networks

KW - Dataset

KW - Free-living

KW - Human Activity Recognition

KW - Machine Learning

UR - http://www.smart-world.org/2019/

M3 - Conference contribution

SP - 166

EP - 171

BT - The 2nd Workshop on advanced Technologies for Smarter Assisted Living solutions: Towards an open Smart Home infrastructure (SmarterAAL'19)

ER -

Cruciani F, Sun C, Zhang S, Nugent CD, Li C, Song S et al. A Public Domain Dataset for Human Activity Recognition in Free-living Conditions. In The 2nd Workshop on advanced Technologies for Smarter Assisted Living solutions: Towards an open Smart Home infrastructure (SmarterAAL'19). 2019. p. 166-171