Related Experiment Videos
Dynamic muscle force predictions from EMG: an artificial neural network approach
M M Liu1, W Herzog, H H Savelberg
1Faculty of Kinesiology, The University of Calgary, Alberta, Canada.
Summary
Artificial neural networks (ANNs) can predict muscle forces from electromyography (EMG) signals across different subjects. This powerful tool accurately captures dynamic muscle contractions, advancing biomechanical research.
Area of Science:
- Biomechanics
- Neuroscience
- Computational Biology
Background:
- Predicting muscle force from electromyography (EMG) signals across subjects is challenging.
- Previous methods have not utilized dynamic EMG signals for cross-subject force prediction.
Purpose of the Study:
- To develop and validate an artificial neural network (ANN) model for predicting muscle forces from EMG signals.
- To assess the ANN model's ability to generalize across different subjects and locomotor conditions.
Main Methods:
- An ANN was trained using a subset of experimentally determined EMG and muscle force data from the cat soleus.
- The trained ANN model was used to predict muscle forces for conditions and subjects not included in the training set.
- Predicted muscle forces were validated against in vivo recorded forces.
Main Results:
- The ANN approach successfully derived an EMG-force relationship, enabling accurate muscle force prediction.
- Intra-subject predictions showed superior results compared to previous studies.
- Inter-subject predictions achieved excellent accuracy, with cross-correlation coefficients >0.90 and root mean square errors <15%.
Conclusions:
- Artificial neural networks are powerful tools for modeling EMG-force relationships in dynamic muscle contractions.
- ANNs can reliably predict muscle forces across subjects, offering a significant advancement in biomechanical analysis.
- This approach holds potential for widespread application in predicting muscle forces from EMG signals.