Evaluation of a long short-term memory (LSTM)-based algorithm for predicting central frequency and synergy activation
Jun Won Choi1, Woon Mo Jung1, Jong Min Kim1
1Department of Biomedical Engineering, Yonsei University, Wonju-Si, Gangwon-Do 26493 Republic of Korea.
Biomedical Engineering Letters
|November 24, 2025
Summary
This study developed a deep learning model using surface electromyography (EMG) and motion analysis to accurately predict muscle fatigue and coordination during exercise. The algorithm shows promise for real-time monitoring in training and rehabilitation.
Area of Science:
- Biomechanics
- Sports Science
- Machine Learning
Background:
- Accurate, non-invasive monitoring of muscle fatigue and coordination is crucial for optimizing exercise performance and rehabilitation.
- Existing methods may be invasive or lack real-time feedback capabilities.
Purpose of the Study:
- To develop and validate a deep learning algorithm for estimating muscle fatigue and intermuscular coordination.
- To integrate surface electromyography (EMG) and markerless motion analysis for comprehensive movement assessment.
Main Methods:
- A deep learning model (LSTM) was trained using synchronized EMG and markerless motion data from participants performing dumbbell curls.
- Muscle fatigue was assessed via median frequency (MDF) of EMG signals.
- Intermuscular coordination was quantified using the Synergy Activation Ratio (SAR).
Main Results:
- The LSTM model achieved high prediction accuracy for both muscle fatigue (MDF) and coordination (SAR).
- Results indicated localized biceps fatigue and potential compensatory activation in the deltoid.
- A decrease in SAR over time suggested fatigue-induced changes in muscle synergy.
Conclusions:
- The proposed deep learning framework effectively detects real-time muscle fatigue and coordination changes.
- This approach holds potential for personalized training, fatigue monitoring, and ergonomic assessments.
