Feature Engineering for Activity Recognition from Wrist-Worn Motion Sensors

Sumeyye Konak, Fulya Turan, M. Shoaib, O. Durmaz

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

Abstract

With their integrated sensors, wrist-worn devices, such as smart watches, provide an ideal platform for human activity recognition. Particularly, the inertial sensors, such as accelerometer and gyroscope can efficiently capture the wrist and arm movements of the users. In this paper, we investigate the use of accelerometer sensor for recognizing thirteen different activities. Particularly, we analyse how different sets of features extracted from acceleration readings perform in activity recognition. We categorize the set of features into three classes: motion related features, orientation-related features and rotation-related features and we analyse the recognition performance using motion, orientation and rotation information both alone and in combination. We utilize a dataset collected from 10 participants and use different classification algorithms in the analysis. The results show that using orientation features achieve the highest accuracies when used alone and in combination wit h other sensors. Moreover, using only raw acceleration performs slightly better than using linear acceleration and similar compared with gyroscope
LanguageUndefined
Title of host publicationProceedings of the 6th International Joint Conference on Pervasive and Embedded Computing and Communication Systems, PECCS 2016
Place of PublicationPortugal
PublisherSCITEPRESS – Science and Technology Publications
Pages76-84
Number of pages9
ISBN (Print)978-989-758-195-3
DOIs
StatePublished - Jul 2016
Event6th International Joint Conference on Pervasive and Embedded Computing and Communication Systems 2016 - Lisbon, Portugal
Duration: 25 Jul 201627 Jul 2016
Conference number: 6
http://www.peccs.org/?y=2016

Publication series

Name
PublisherSCITEPRESS – Science and Technology Publications

Conference

Conference6th International Joint Conference on Pervasive and Embedded Computing and Communication Systems 2016
Abbreviated titlePECCS 2016
CountryPortugal
CityLisbon
Period25/07/1627/07/16
OtherPECCS 2016 will be held in conjunction with ICETE 2016 and PhyCS 2016
Internet address

Keywords

  • EWI-27123
  • IR-101362
  • METIS-318485

Cite this

Konak, S., Turan, F., Shoaib, M., & Durmaz, O. (2016). Feature Engineering for Activity Recognition from Wrist-Worn Motion Sensors. In Proceedings of the 6th International Joint Conference on Pervasive and Embedded Computing and Communication Systems, PECCS 2016 (pp. 76-84). Portugal: SCITEPRESS – Science and Technology Publications. DOI: 10.5220/0006007100760084
Konak, Sumeyye ; Turan, Fulya ; Shoaib, M. ; Durmaz, O./ Feature Engineering for Activity Recognition from Wrist-Worn Motion Sensors. Proceedings of the 6th International Joint Conference on Pervasive and Embedded Computing and Communication Systems, PECCS 2016. Portugal : SCITEPRESS – Science and Technology Publications, 2016. pp. 76-84
@inproceedings{b3cf128b85064d66aae0e7336f69e39b,
title = "Feature Engineering for Activity Recognition from Wrist-Worn Motion Sensors",
abstract = "With their integrated sensors, wrist-worn devices, such as smart watches, provide an ideal platform for human activity recognition. Particularly, the inertial sensors, such as accelerometer and gyroscope can efficiently capture the wrist and arm movements of the users. In this paper, we investigate the use of accelerometer sensor for recognizing thirteen different activities. Particularly, we analyse how different sets of features extracted from acceleration readings perform in activity recognition. We categorize the set of features into three classes: motion related features, orientation-related features and rotation-related features and we analyse the recognition performance using motion, orientation and rotation information both alone and in combination. We utilize a dataset collected from 10 participants and use different classification algorithms in the analysis. The results show that using orientation features achieve the highest accuracies when used alone and in combination wit h other sensors. Moreover, using only raw acceleration performs slightly better than using linear acceleration and similar compared with gyroscope",
keywords = "EWI-27123, IR-101362, METIS-318485",
author = "Sumeyye Konak and Fulya Turan and M. Shoaib and O. Durmaz",
note = "eemcs-eprint-27123",
year = "2016",
month = "7",
doi = "10.5220/0006007100760084",
language = "Undefined",
isbn = "978-989-758-195-3",
publisher = "SCITEPRESS – Science and Technology Publications",
pages = "76--84",
booktitle = "Proceedings of the 6th International Joint Conference on Pervasive and Embedded Computing and Communication Systems, PECCS 2016",

}

Konak, S, Turan, F, Shoaib, M & Durmaz, O 2016, Feature Engineering for Activity Recognition from Wrist-Worn Motion Sensors. in Proceedings of the 6th International Joint Conference on Pervasive and Embedded Computing and Communication Systems, PECCS 2016. SCITEPRESS – Science and Technology Publications, Portugal, pp. 76-84, 6th International Joint Conference on Pervasive and Embedded Computing and Communication Systems 2016, Lisbon, Portugal, 25/07/16. DOI: 10.5220/0006007100760084

Feature Engineering for Activity Recognition from Wrist-Worn Motion Sensors. / Konak, Sumeyye; Turan, Fulya; Shoaib, M.; Durmaz, O.

Proceedings of the 6th International Joint Conference on Pervasive and Embedded Computing and Communication Systems, PECCS 2016. Portugal : SCITEPRESS – Science and Technology Publications, 2016. p. 76-84.

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

TY - GEN

T1 - Feature Engineering for Activity Recognition from Wrist-Worn Motion Sensors

AU - Konak,Sumeyye

AU - Turan,Fulya

AU - Shoaib,M.

AU - Durmaz,O.

N1 - eemcs-eprint-27123

PY - 2016/7

Y1 - 2016/7

N2 - With their integrated sensors, wrist-worn devices, such as smart watches, provide an ideal platform for human activity recognition. Particularly, the inertial sensors, such as accelerometer and gyroscope can efficiently capture the wrist and arm movements of the users. In this paper, we investigate the use of accelerometer sensor for recognizing thirteen different activities. Particularly, we analyse how different sets of features extracted from acceleration readings perform in activity recognition. We categorize the set of features into three classes: motion related features, orientation-related features and rotation-related features and we analyse the recognition performance using motion, orientation and rotation information both alone and in combination. We utilize a dataset collected from 10 participants and use different classification algorithms in the analysis. The results show that using orientation features achieve the highest accuracies when used alone and in combination wit h other sensors. Moreover, using only raw acceleration performs slightly better than using linear acceleration and similar compared with gyroscope

AB - With their integrated sensors, wrist-worn devices, such as smart watches, provide an ideal platform for human activity recognition. Particularly, the inertial sensors, such as accelerometer and gyroscope can efficiently capture the wrist and arm movements of the users. In this paper, we investigate the use of accelerometer sensor for recognizing thirteen different activities. Particularly, we analyse how different sets of features extracted from acceleration readings perform in activity recognition. We categorize the set of features into three classes: motion related features, orientation-related features and rotation-related features and we analyse the recognition performance using motion, orientation and rotation information both alone and in combination. We utilize a dataset collected from 10 participants and use different classification algorithms in the analysis. The results show that using orientation features achieve the highest accuracies when used alone and in combination wit h other sensors. Moreover, using only raw acceleration performs slightly better than using linear acceleration and similar compared with gyroscope

KW - EWI-27123

KW - IR-101362

KW - METIS-318485

U2 - 10.5220/0006007100760084

DO - 10.5220/0006007100760084

M3 - Conference contribution

SN - 978-989-758-195-3

SP - 76

EP - 84

BT - Proceedings of the 6th International Joint Conference on Pervasive and Embedded Computing and Communication Systems, PECCS 2016

PB - SCITEPRESS – Science and Technology Publications

CY - Portugal

ER -

Konak S, Turan F, Shoaib M, Durmaz O. Feature Engineering for Activity Recognition from Wrist-Worn Motion Sensors. In Proceedings of the 6th International Joint Conference on Pervasive and Embedded Computing and Communication Systems, PECCS 2016. Portugal: SCITEPRESS – Science and Technology Publications. 2016. p. 76-84. Available from, DOI: 10.5220/0006007100760084