Koki Kumura, Tamotsu Kamigaki
In manufacturing plants, Industrial Engineering (IE) methods have been applied to reduce workers' burden and improve productivity by addressing overburden, waste, and unevenness. Representative approaches include the OWAS method, which evaluates physical load by observing workers' postures, Sabrick analysis, which records each task using symbols and diagrams, and the stopwatch method, which measures task duration. Although video recording devices allow observation without being on site, the identification of overburden, waste, and unevenness still requires detailed analysis of recorded data, demanding considerable effort. To simplify this process, various motion analysis techniques using specialized equipment have been studied; however, their high-cost limits practical adoption. To address this issue, this study proposes a low-cost and user-friendly motion analysis method using mocopi®, a mobile motion capture system. We developed a system that captures workers' movements, visualizes them in Unity, and outputs coordinate data. Furthermore, by applying for a Support Vector Machine (SVM), we constructed a posture classification model that successfully distinguished four actions: standing, squatting, hand movements, and forward bending. Future work will focus on enabling real-time classification.