Machine Learning Physical Fatigue Estimation Approach Based on IMU and EMG Wearable Sensors

Physical fatigue refers to a state of exhaustion or reduced capacity for physical performance due to prolonged exertion, repetitive movements, or lack of rest. It is a multifaceted condition that can severely impact performance, especially in activities requiring sustained effort, precision, or concentration. In physical tasks, fatigue manifests as a decrease in muscle strength, coordination, and endurance, leading to diminished performance and an increased risk of injury. Detecting physical fatigue is crucial in a variety of domains: professional sports, collaborative robotics, construction, and more. This research introduces a novel framework for predicting fatigue during shoulder movements using data collected from wearable inertial measurement units and electromyography sensors. By integrating the Borg Scale, a subjective measure of perceived exertion, our approach uniquely combines objective sensor data with user-reported fatigue levels, creating a more holistic fatigue assessment model. The primary aim of this study is to develop a predictive model capable of accurately estimating fatigue, as measured by the Borg Scale. An investigation of the best machine learning algorithm for this task ensures that the chosen method provides the most reliable predictions. Furthermore, by systematically reducing the number of sensors and analyzing the impact on model performance, it is possible to find a minimal sensor configuration that maintains the model’s predictive power while reducing complexity and cost. The Ridge Regression model, after hyperparameter tuning, outperformed other models, achieving a mean absolute error of 2.417 in predicting fatigue. This preliminary study shows the potential of integrating data from different inertial and electromyography sensors for fatigue prediction in shoulder movements, with potential applications in occupational safety.