An Improved Compound Gaussian Model for Bivariate Surface EMG Signals Related to Strength Training

TL;DR

This paper introduces an improved compound Gaussian model for multivariate surface EMG signals that better captures their non-stationary statistical properties, validated on a new dataset and linked to training variables.

Contribution

The study proposes a novel compound Gaussian model with an exponential latent covariance variable for sEMG signals, enhancing statistical fit over existing models.

Findings

The proposed model fits empirical sEMG data more closely than previous models.

Model parameters correlate with training weights and activity types.

Enhanced statistical modeling improves understanding of sEMG signal behavior.

Abstract

Recent literature suggests that the surface electromyography (sEMG) signals have non-stationary statistical characteristics specifically due to random nature of the covariance. Thus suitability of a statistical model for sEMG signals is determined by the choice of an appropriate model for describing the covariance. The purpose of this study is to propose a Compound-Gaussian (CG) model for multivariate sEMG signals in which latent variable of covariance is modeled as a random variable that follows an exponential model. The parameters of the model are estimated using the iterative Expectation Maximization (EM) algorithm. Further, a new dataset, electromyography analysis of human activities database 2 (EMAHA-DB2) is developed. Based on the model fitting analysis on the sEMG signals from EMAHA-DB2, it is found that the proposed CG model fits more closely to the empirical pdf of sEMG signals than the existing models. The proposed model is validated by visual inspection, further validated by matching central moments and better quantitative metrics in comparison with other models. The proposed compound model provides an improved fit to the statistical behavior of sEMG signals. Further, the estimate of rate parameter of the exponential model shows clear relation to the training weights. Finally, the average signal power estimates of the channels shows distinctive dependency on the training weights, the subject's training experience and the type of activity.