The Estimation of Sagittal Plane Shoulder Flexion/Extension Posture Endurance Time and Repetition Endurance time on Drilling Tasks from Surface Electromyography (sEMG) Signals.

This paper explores the global issue of Work-Related Musculoskeletal Disorders (WMSDs), impacting around 1.71 billion people worldwide. WMSDs, causing pain primarily in the neck, shoulders, and lower back, not only affect individual workers negatively but also result in increased productivity losses and societal costs. Current prevention strategies involve subjective surveys and ergonomic assessments using tools like RULA, REBA, and NIOSH lifting equation. However, these methods have limitations in pinpointing specific improvements for tasks, as evident in cases where high-risk tasks, according to assessments, didn't align with workers' experiences. To address this, the study proposes a foundational exploration for developing a digital musculoskeletal workload assessment system using deep neural networks. The focus is on classifying shoulder flexion/extension postures and estimating endurance times during drilling tasks through surface electromyography (sEMG) signals. The methodology involves data collection from healthy male adults performing shoulder-related tasks, utilizing sEMG sensors. The collected data is processed, and a deep neural network model, incorporating recurrent neural networks (RNNs), is developed for posture and task classification as well as endurance time estimation. The study acknowledges the need for further refinement in hyperparameter tuning and feature engineering to enhance model performance, aiming to contribute valuable insights for improving industrial work environments and preventing musculoskeletal injuries.