TY - GEN
T1 - PoseX
T2 - 2020 IEEE EMBS Conference on Biomedical Engineering and Sciences, IECBES 2020
AU - Wu, Bing Fei
AU - Lin, Chien Chou
AU - Huang, Po Wei
N1 - Publisher Copyright:
© 2021 IEEE.
PY - 2021/3/1
Y1 - 2021/3/1
N2 - As computers are widely used in daily life, computer users spend a lot of time staring at screens. long-term sitting postures lead to upper body pain which brings negative physical and mental impacts. Any exaggerated forward curvature in the upper back is called kyphosis. The common unhealthy postures are cervical kyphosis, thoracic kyphosis, and raised-shoulders. PoseX can recognize cervical spine kyphosis, thoracic spine kyphosis, and raised-shoulders from a facial image with only one general webcam. Based on the body key-points provided by OpenPose, PoseX uses the proportional relation between face width, shoulders width, and ears-shoulders height as the features to extract depth information from the RGB image. In our experiment, 15 computer users were recruited. The sitting scenarios include four tasks (normal, cervical kyphosis, thoracic kyphosis, and raised-shoulders) and three distance conditions (60 cm, 85 cm, 110 cm), PoseX records personalized parameters in 3 seconds and performs an accuracy higher or equal to 84 percent in each task. The proposed technique shows a promising low-cost health monitoring system for computer users.
AB - As computers are widely used in daily life, computer users spend a lot of time staring at screens. long-term sitting postures lead to upper body pain which brings negative physical and mental impacts. Any exaggerated forward curvature in the upper back is called kyphosis. The common unhealthy postures are cervical kyphosis, thoracic kyphosis, and raised-shoulders. PoseX can recognize cervical spine kyphosis, thoracic spine kyphosis, and raised-shoulders from a facial image with only one general webcam. Based on the body key-points provided by OpenPose, PoseX uses the proportional relation between face width, shoulders width, and ears-shoulders height as the features to extract depth information from the RGB image. In our experiment, 15 computer users were recruited. The sitting scenarios include four tasks (normal, cervical kyphosis, thoracic kyphosis, and raised-shoulders) and three distance conditions (60 cm, 85 cm, 110 cm), PoseX records personalized parameters in 3 seconds and performs an accuracy higher or equal to 84 percent in each task. The proposed technique shows a promising low-cost health monitoring system for computer users.
KW - Kyphosis
KW - postural syndromes
KW - posture detection
KW - sensing system
KW - webcam-based
UR - http://www.scopus.com/inward/record.url?scp=85104829212&partnerID=8YFLogxK
U2 - 10.1109/IECBES48179.2021.9398773
DO - 10.1109/IECBES48179.2021.9398773
M3 - Conference contribution
AN - SCOPUS:85104829212
T3 - Proceedings - 2020 IEEE EMBS Conference on Biomedical Engineering and Sciences, IECBES 2020
SP - 109
EP - 114
BT - Proceedings - 2020 IEEE EMBS Conference on Biomedical Engineering and Sciences, IECBES 2020
PB - Institute of Electrical and Electronics Engineers Inc.
Y2 - 1 March 2021 through 3 March 2021
ER -