Novel Wearable Monitoring System of Forward Head Posture Assisted by Magnet-Magnetometer Pair and Machine Learning
- Authors
- Han, Hobeom; Jang, Hyeongkyu; Yoon, Sang Won
- Issue Date
- Apr-2020
- Publisher
- IEEE-INST ELECTRICAL ELECTRONICS ENGINEERS INC
- Keywords
- Forward head posture; magnetometer; machine learning; wearable monitoring
- Citation
- IEEE SENSORS JOURNAL, v.20, no.7, pp.3838 - 3848
- Indexed
- SCIE
SCOPUS
- Journal Title
- IEEE SENSORS JOURNAL
- Volume
- 20
- Number
- 7
- Start Page
- 3838
- End Page
- 3848
- URI
- https://scholarworks.bwise.kr/hanyang/handle/2021.sw.hanyang/3735
- DOI
- 10.1109/JSEN.2019.2959817
- ISSN
- 1530-437X
- Abstract
- Forward head posture (FHP) increasingly threatens human health through numerous disorders. Wearable sensing methods facilitate promising solutions to portably quantify FHP symptoms. Flex and acceleration sensors are two major wearable sensors; however, several technical challenges still remain. This work proposes a novel wearable FHP sensing solution relying on a three-axis magnetometer paired with a miniature permanent magnet. The magnetometer precisely calibrates head postures by tracking directional changes of the magnetic field from the magnet and by sensor-fusing with an accelerometer. Sensor-fused data are processed by machine learning algorithms, either to provide neck-angle values representing craniovertebral angle with reduced noise (regression algorithms) or to determine risk levels of FHP (classification algorithms). Performances of four regression and four classification algorithms are compared for two experimental scenarios (named calibration-mode and usage-mode scenarios). In both scenarios, our sensor-fusion with the magnet-magnetometer pair exhibited outstanding performances compared to a conventional accelerometer approach. The performances differed by machine learning algorithms and scenarios, but reliably demonstrated extremely high correlation coefficients (R =0.9945) and classification accuracy (similar to 95.6%) in the calibration-mode scenario. When participants watched videos using their own smartphones (usage mode), a high correlation coefficient (R =similar to 0.9365) and classification accuracy (similar to>90.4%) were achieved.
- Files in This Item
-
Go to Link
- Appears in
Collections - 서울 공과대학 > 서울 미래자동차공학과 > 1. Journal Articles
Items in ScholarWorks are protected by copyright, with all rights reserved, unless otherwise indicated.