3D human pose keypoints and corresponding joint angle calculation for vision-based WMSD risk assessments
Leyang Wen et al.
What the paper says
Vision-based pose estimation, which can utilise ordinary videos, has been applied to assess work-related musculoskeletal disorder (WMSD) risks as a less-intrusive and accessible method. However, the underlying machine learning models are built from generic pose datasets, lacking critical 3D information for calculating high-degree-of-freedom joint angles needed in WMSD risk assessments. We defined a 66-keypoint set specialised for joint angle calculations while containing visual features suitable for vision-based pose estimation and derived corresponding angle calculation steps. To test its usefulness, we collected a 6.7-million-frame dataset featuring 9 categories of manual material handling and assembly tasks to train a baseline pose estimation model for joint angle calculations. Our approach enabled the calculation of 22 angles across 6 body joints for WMSD risk assessments with a mean absolute angle error of 2.4° when a baseline pose estimation model was used, which demonstrates its usefulness for joint angle calculations for WMSD risk assessments.
Evidence weight
Balanced mode · F 0.40 / M 0.15 / V 0.05 / R 0.40
| F · citation impact | 0.50 × 0.4 = 0.20 |
| M · momentum | 0.50 × 0.15 = 0.07 |
| V · venue signal | 0.50 × 0.05 = 0.03 |
| R · text relevance † | 0.50 × 0.4 = 0.20 |
† Text relevance is estimated at 0.50 on the detail page — for your query’s actual relevance score, open this paper from a search result.