Gaussian-Laplacian Mixture-Enhanced AGLR for Accurate EMG Onset Detection.

Area of Science:

• Biomedical Engineering
• Signal Processing
• Neuroscience

Background:

• Automated detection of muscle contraction onset in electromyography (EMG) signals is vital for clinical applications and human-machine interfaces.
• Existing methods often assume Gaussian signal characteristics, which are inaccurate for submaximal contractions exhibiting Laplacian-like properties.

Purpose of the Study:

• To introduce a novel algorithm, GLM-AGLR (Gaussian-Laplacian mixture model based approximated generalized likelihood ratio), for improved EMG onset detection.
• To address the limitations of Gaussian-based methods in accurately detecting onsets during submaximal muscle contractions.

Main Methods:

• Developed a Gaussian-Laplacian mixture model to represent the muscle contraction phase in EMG signals.
• Implemented an approximated generalized likelihood ratio (AGLR) framework incorporating the mixture model.
• Validated the GLM-AGLR algorithm using EMG data collected at various contraction intensities.

Main Results:

• The GLM-AGLR algorithm demonstrated superior accuracy in detecting EMG onset compared to conventional threshold-based methods and prior AGLR variants.
• Results closely matched visually identified ground truth, indicating enhanced reliability.
• The model effectively handles EMG signals with both Gaussian and Laplacian characteristics.

Conclusions:

• The proposed GLM-AGLR algorithm offers a more robust approach to automated EMG onset detection, particularly for submaximal contractions.
• This advancement improves EMG signal segmentation and feature extraction, supporting more reliable EMG-based technologies.
• Enhanced detection accuracy has significant clinical relevance for diagnostics and patient-focused applications.