Presentation
Data-Driven Classification of Manual Material Handling Tasks through Markerless Motion Capture Using Recurrent Neural Networks
DescriptionManual material handling tasks are associated with multiple risk factors for work-related musculoskeletal disorders (WMSDs) and substantially contribute to the high prevalence of WMSDs. Accurate and efficient physical exposure assessment methods are needed to quantify WMSD risks and establish the relationship between exposures and risks. While several tools are available, they are often limited in scope and can be resource intensive. We investigated the use of kinematic data from a markerless motion capture system, together with various machine-learning algorithms and input data features, to classify among eight distinct material handling tasks. Using kinematic data led to satisfactory results (e.g., mean precision of 85 – 97%) in classifying the tasks and the task conditions. Our results, though, also emphasize that classification performance differed between machine learning algorithms, feature sets, tasks, and between males and females. Nonetheless, the use of markerless motion capture appears to have clear potential for physical exposure assessment.
Contributors
Research Assistant Professor
H.G. Prillaman Professor, Assistant Department Head, and Graduate Program Director
Event Type
Lecture
TimeWednesday, September 11th3pm - 3:20pm MST
LocationFLW Salon H
Occupational Ergonomics