Comparison of Signal Processing Methods for Reducing Motion Artifacts in High-Density Electromyography During Human
Bryan R Schlink1, Andrew D Nordin1, Daniel P Ferris1
1J. Crayton Pruitt Family Department of Biomedical EngineeringUniversity of Florida Gainesville FL 32608 USA.
IEEE Open Journal of Engineering in Medicine and Biology
|April 11, 2022
Summary
Canonical correlation analysis effectively removes motion artifacts from high-density electromyography (EMG) during locomotion. This method outperforms traditional filtering for clearer myoelectric signal analysis in running and walking.
Area of Science:
- Biomechanics
- Neuroscience
- Signal Processing
Background:
- High-density electromyography (EMG) is crucial for analyzing muscle activity during human movement.
- Locomotion, such as walking and running, introduces motion artifacts that contaminate EMG signals.
- Existing filtering methods may not adequately remove these artifacts without compromising the myoelectric signal.
Purpose of the Study:
- To compare the efficacy of canonical correlation analysis (CCA) and principal component analysis (PCA) against traditional high-pass filtering for motion artifact removal.
- To evaluate the performance of these methods on high-density EMG data from the gastrocnemius and tibialis anterior muscles.
- To determine the optimal signal decomposition and filtering approach for high-density EMG during locomotion.
Main Methods:
- High-density EMG data were collected from gastrocnemius and tibialis anterior muscles during walking and running.
- Signals were processed using traditional high-pass filtering, PCA-based filtering, and CCA-based filtering.
- Performance was quantified by assessing the reduction of motion artifact frequencies and preservation of myoelectric signal frequencies.
Main Results:
- CCA filtering demonstrated superior reduction of signal content in frequency bands associated with motion artifacts compared to high-pass and PCA filtering.
- CCA filtering minimized the reduction of signal content in frequency bands characteristic of true myoelectric signals.
- Both walking and running data showed improved artifact removal with CCA.
Conclusions:
- Canonical correlation analysis filtering is a more effective technique for cleaning high-density EMG signals contaminated by motion artifacts during fast walking and running.
- CCA preserves the integrity of the myoelectric signal better than traditional high-pass or PCA filtering methods.
- This advancement offers improved accuracy for biomechanical and neurophysiological studies involving locomotion.
