Personalized EMG preprocessing and normalization for musculoskeletal simulation
Dovydas Cicėnas1, Jurgita Žižienė1, Kristina Daunoravičienė1
1Department of Biomechanical Engineering, Faculty of Mechanics, Vilnius Gediminas Technical University, Vilnius, Lithuania.
Summary
This study developed a personalized electromyography (EMG) preprocessing pipeline to improve the accuracy of musculoskeletal (MS) models for biomechanics. The adaptive approach enhances physiological realism in EMG-driven models for rehabilitation applications.
Area of Science:
- Biomechanics
- Rehabilitation Engineering
- Signal Processing
Background:
- Accurate electromyography (EMG) signal interpretation is crucial for musculoskeletal (MS) models in biomechanics and rehabilitation.
- Conventional preprocessing methods often fail to capture subject-specific signal characteristics and task-specific muscle functions.
Purpose of the Study:
- To develop and validate an adaptive, personalized EMG preprocessing pipeline.
- To enhance the physiological accuracy of EMG-driven MS models during elbow flexion-extension tasks.
Main Methods:
- Recorded EMG signals from six upper limb muscles during elbow movements.
- Applied individualized spectral filtering and a dual-stage normalization (dynamic MVC-based min-max, functional weighting).
- Validated the processed EMG signals using an OpenSim elbow model against Xsens motion capture data.
Main Results:
- The EMG-driven OpenSim model demonstrated strong agreement with reference motion data (R > 0.98, RMSE < 8°).
- Joint angle trajectories were consistent and physiologically plausible, despite a minor systematic offset.
- The subject-specific pipeline improved model accuracy and interpretability.
Conclusions:
- The proposed subject-specific EMG preprocessing pipeline enhances biomechanical model accuracy and interpretability.
- Future work should investigate adaptive signal alignment and AI for improved robustness in dynamic and wearable scenarios.
