Evaluation of muscle force classification using shape analysis of the sEMG probability density function: a simulation...

Area of Science:

• Biomedical Engineering
• Neuroscience
• Signal Processing

Background:

• Surface electromyogram (sEMG) amplitude probability density function (PDF) shape is influenced by contraction level, fatigue, muscle anatomy, instrumentation, and neural control.
• Understanding sEMG PDF shape variability is crucial for accurately assessing neural drive and muscle activation strategies.

Purpose of the Study:

• To classify the shape variability of the sEMG amplitude PDF across different contraction levels using high-order statistics (HOS) and Core Shape Modeling (CSM).
• To evaluate the sensitivity of shape analysis methods to physiological, instrumental, and neural control parameters.
• To compare the performance of CSM and HOS with classical amplitude estimators (ARV, RMS) in classifying sEMG data.

Main Methods:

• Large-scale simulation using an sEMG-force model and parallel computing, incorporating 25 muscle anatomies, 10 parameter configurations, and 3 electrode arrangements.
• Classification of sEMG data from three contraction levels (20%, 50%, 80% MVC).
• Application of a shape clustering algorithm using five HOS combinations and CSM, compared against ARV and RMS amplitude clustering.

Main Results:

• The CSM method, particularly with a Laplacian electrode arrangement, achieved high classification scores, comparable to ARV and RMS, and superior to some HOS combinations.
• Classification scores decreased when critical confounding parameters were altered, highlighting the sensitivity of shape analysis.
• The study confirmed that sEMG amplitude PDF shape analysis is complex and requires robust methods and specific recording protocols.

Conclusions:

• Core Shape Modeling (CSM) and high-order statistics (HOS) offer valuable, albeit complex, methods for analyzing sEMG amplitude PDF shape variability.
• Accurate tracking of neural drive and muscle activation strategies necessitates efficient shape analysis techniques and optimized signal recording protocols.
• Classical amplitude estimators (ARV, RMS) remain effective, but shape analysis provides complementary insights, especially when combined with advanced methods and careful experimental design.