A Laplacian-Gaussian Mixture Model for Surface EMG Signals from Upper Limbs

Area of Science:

• Biomedical Engineering
• Signal Processing
• Computational Neuroscience

Background:

• Surface Electromyography (sEMG) signal probability density functions (pdf) are often modeled as Gaussian or Laplacian.
• Accurate modeling of sEMG signals is crucial for understanding muscle activity and developing advanced prosthetics or control systems.

Purpose of the Study:

• To propose and validate a novel Laplacian-Gaussian mixture model for sEMG signals from upper limbs.
• To assess the contribution of Laplacian and Gaussian components in modeling sEMG signals.

Main Methods:

• Developed a Laplacian-Gaussian mixture model for sEMG signal analysis.
• Employed the Expectation-Maximization (EM) algorithm for parameter estimation.
• Validated the model using Kullback-Leibler (KL) divergence and goodness-of-fit plots (R-squared).

Main Results:

• The proposed Laplacian-Gaussian mixture model demonstrated strong quantitative and qualitative agreement with empirical sEMG probability density functions.
• The Kullback-Leibler (KL) divergence confirmed the model's accuracy against benchmark datasets.
• Analysis revealed a significant contribution of the Laplacian component to the overall sEMG signal mixture.

Conclusions:

• The Laplacian-Gaussian mixture model provides a more accurate representation of upper limb sEMG signals compared to traditional models.
• The significant weight of the Laplacian component highlights its importance in characterizing sEMG signal dynamics.
• This refined model can enhance applications relying on precise sEMG signal interpretation.