Presentation
Sensors and algorithms for automated prediction threshold limit values for worker hand activity
DescriptionManual material handling tasks remain common in industrial settings and can expose workers to hand-intensive risk factors associated with work-related musculoskeletal disorders. This study developed and evaluated a wearable-sensor approach for estimating the American Conference of Governmental Industrial Hygienists Threshold Limit Value (TLV®) for Hand Activity Level (HAL). Twelve participants performed simulated material-transfer tasks across systematically varied hand activity and force conditions. During each task, participants wore a pressure-sensing tactile glove and a wrist-mounted inertial measurement unit (IMU). Glove data were summarized into force features, including mean force, peak force, and force-gradient measures, while IMU data were used to derive kinematic features from linear acceleration and angular velocity. HAL and normalized peak force (NPF) were modeled using linear mixed-effects models with subject-specific random intercepts. Predicted HAL and NPF were then used to compute estimated TLV® values. IMU-derived motion intensity feature strongly predicted HAL, achieving a marginal R² of 0.866. Glove-derived force features moderately predicted NPF, with a conditional R² of 0.433. The final predicted TLV® values showed moderate agreement with ground truth, with RMSE = 0.319 and Spearman ρ = 0.466. These findings suggest that wearable tactile and motion sensors may support automated ergonomic assessment for hand-intensive manual tasks.
Event Type
Lecture
TimeWednesday, October 21st5:10pm - 5:30pm PDT
Location