Left-right swapping and upper-lower limb pairing for robust multi-wearable workout activity detection
- Author
- Jonas Van Der Donckt (UGent) , Jeroen Van Der Donckt (UGent) and Sofie Van Hoecke (UGent)
- Organization
- Project
- Abstract
- This work presents the solution of the Signal Sleuths team for the 2024 HASCA WEAR challenge. The challenge focuses on detecting 18 workout activities (and the null class) using accelerometer data from 4 wearables - one worn on each limb. Data analysis revealed inconsistencies in wearable orientation within and across participants, leading to exploring novel multi-wearable data augmentation techniques. We investigate three models using a fixed feature set: (i) "raw": using all data as is, (ii) "left-right swapping": augmenting data by swapping left and right limb pairs, and (iii) "upper-lower limb paring": stacking data by using upper-lower limb pair combinations (2 wearables). Our experiments utilize traditional machine learning with multi-window feature extraction and temporal smoothing. Using 3-fold cross-validation, the raw model achieves a macro F1-score of 90.01%, whereas left-right swapping and upper-lower limb paring improve the scores to 91.30% and 91.87% respectively.
- Keywords
- Machine Learning, Human Activity Recognition, Wearables, Multimodal Sensors, WEAR Dataset
Downloads
-
(...).pdf
- full text (Published version)
- |
- UGent only
- |
- |
- 1.47 MB
Citation
Please use this url to cite or link to this publication: http://hdl.handle.net/1854/LU-01JBVAWCHASA22M6MGZV0M42HT
- MLA
- Van Der Donckt, Jonas, et al. “Left-Right Swapping and Upper-Lower Limb Pairing for Robust Multi-Wearable Workout Activity Detection.” COMPANION OF THE 2024 ACM INTERNATIONAL JOINT CONFERENCE ON PERVASIVE AND UBIQUITOUS COMPUTING, UBICOMP COMPANION 2024, 2024, pp. 545–50, doi:10.1145/3675094.3678453.
- APA
- Van Der Donckt, J., Van Der Donckt, J., & Van Hoecke, S. (2024). Left-right swapping and upper-lower limb pairing for robust multi-wearable workout activity detection. COMPANION OF THE 2024 ACM INTERNATIONAL JOINT CONFERENCE ON PERVASIVE AND UBIQUITOUS COMPUTING, UBICOMP COMPANION 2024, 545–550. https://doi.org/10.1145/3675094.3678453
- Chicago author-date
- Van Der Donckt, Jonas, Jeroen Van Der Donckt, and Sofie Van Hoecke. 2024. “Left-Right Swapping and Upper-Lower Limb Pairing for Robust Multi-Wearable Workout Activity Detection.” In COMPANION OF THE 2024 ACM INTERNATIONAL JOINT CONFERENCE ON PERVASIVE AND UBIQUITOUS COMPUTING, UBICOMP COMPANION 2024, 545–50. https://doi.org/10.1145/3675094.3678453.
- Chicago author-date (all authors)
- Van Der Donckt, Jonas, Jeroen Van Der Donckt, and Sofie Van Hoecke. 2024. “Left-Right Swapping and Upper-Lower Limb Pairing for Robust Multi-Wearable Workout Activity Detection.” In COMPANION OF THE 2024 ACM INTERNATIONAL JOINT CONFERENCE ON PERVASIVE AND UBIQUITOUS COMPUTING, UBICOMP COMPANION 2024, 545–550. doi:10.1145/3675094.3678453.
- Vancouver
- 1.Van Der Donckt J, Van Der Donckt J, Van Hoecke S. Left-right swapping and upper-lower limb pairing for robust multi-wearable workout activity detection. In: COMPANION OF THE 2024 ACM INTERNATIONAL JOINT CONFERENCE ON PERVASIVE AND UBIQUITOUS COMPUTING, UBICOMP COMPANION 2024. 2024. p. 545–50.
- IEEE
- [1]J. Van Der Donckt, J. Van Der Donckt, and S. Van Hoecke, “Left-right swapping and upper-lower limb pairing for robust multi-wearable workout activity detection,” in COMPANION OF THE 2024 ACM INTERNATIONAL JOINT CONFERENCE ON PERVASIVE AND UBIQUITOUS COMPUTING, UBICOMP COMPANION 2024, Melbourne, AUSTRALIA, 2024, pp. 545–550.
@inproceedings{01JBVAWCHASA22M6MGZV0M42HT,
abstract = {{This work presents the solution of the Signal Sleuths team for the 2024 HASCA WEAR challenge. The challenge focuses on detecting 18 workout activities (and the null class) using accelerometer data from 4 wearables - one worn on each limb. Data analysis revealed inconsistencies in wearable orientation within and across participants, leading to exploring novel multi-wearable data augmentation techniques. We investigate three models using a fixed feature set: (i) "raw": using all data as is, (ii) "left-right swapping": augmenting data by swapping left and right limb pairs, and (iii) "upper-lower limb paring": stacking data by using upper-lower limb pair combinations (2 wearables). Our experiments utilize traditional machine learning with multi-window feature extraction and temporal smoothing. Using 3-fold cross-validation, the raw model achieves a macro F1-score of 90.01%, whereas left-right swapping and upper-lower limb paring improve the scores to 91.30% and 91.87% respectively.}},
author = {{Van Der Donckt, Jonas and Van Der Donckt, Jeroen and Van Hoecke, Sofie}},
booktitle = {{COMPANION OF THE 2024 ACM INTERNATIONAL JOINT CONFERENCE ON PERVASIVE AND UBIQUITOUS COMPUTING, UBICOMP COMPANION 2024}},
isbn = {{9798400710582}},
keywords = {{Machine Learning,Human Activity Recognition,Wearables,Multimodal Sensors,WEAR Dataset}},
language = {{eng}},
location = {{Melbourne, AUSTRALIA}},
pages = {{545--550}},
title = {{Left-right swapping and upper-lower limb pairing for robust multi-wearable workout activity detection}},
url = {{http://doi.org/10.1145/3675094.3678453}},
year = {{2024}},
}
- Altmetric
- View in Altmetric
- Web of Science
- Times cited: