Variance distribution analysis of surface EMG signals based on marginal maximum likelihood estimation

Area of Science:

• Biomedical Engineering
• Signal Processing
• Neuroscience

Background:

• Surface electromyogram (EMG) signals are crucial for understanding muscle activity.
• Accurate modeling of EMG signal variance is essential for reliable analysis.
• Existing models may not fully capture the complexities of signal-dependent noise in EMG.

Purpose of the Study:

• To develop and validate a stochastic model for estimating the variance distribution of EMG signals.
• To analyze the relationship between muscle force and EMG variance distribution, considering signal-dependent noise.

Main Methods:

• Assumed Gaussian distribution for EMG signals at a given time.
• Modeled signal variance as a random variable following an inverse gamma distribution.
• Employed marginal likelihood maximization for variance distribution estimation.
• Validated the method using simulated and measured EMG signals.

Main Results:

• The proposed method accurately estimates EMG variance distribution across various shaping parameters.
• Simulation experiments confirmed the high accuracy of the variance distribution estimation.
• Analysis of measured EMG data revealed a significant relationship between muscle force and variance distribution, highlighting the role of signal-dependent noise.

Conclusions:

• The stochastic EMG model provides a robust framework for variance distribution estimation.
• The findings enhance our understanding of muscle force dynamics and EMG signal characteristics.
• This method offers improved analytical capabilities for EMG-based applications.