Clustering analysis and pattern discrimination of EMG linear envelopes.

Area of Science:

• Biomechanics
• Biomedical Engineering
• Signal Processing

Background:

• Electromyography (EMG) analysis is crucial for understanding locomotion and diagnosing injuries.
• Characterizing EMG signal patterns, specifically linear envelopes (LEs), aids in gait analysis.
• Existing methods may not fully capture the dynamic nuances of EMG during movement.

Purpose of the Study:

• To develop and validate a novel technique for pattern analysis of EMG signals during locomotion.
• To determine the key spectral components influencing EMG linear envelope shapes.
• To apply this technique for differentiating gait patterns between healthy and anterior cruciate ligament (ACL) injured individuals.

Main Methods:

• Developed a pattern analysis technique for EMG linear envelopes (LEs) during locomotion.
• Compared Autoregressive (AR) parametric models with Discrete Fourier Transform (DFT) approaches for LE description.
• Employed DFT for feature extraction, focusing on phase spectra and incorporating magnitude spectra and percent powers for harmonic weighting.

Main Results:

• EMG LE shapes are primarily determined by their phase spectra, with magnitude spectra playing a lesser role.
• The DFT approach proved superior to AR models in describing EMG LEs.
• Successful application of the technique for clustering EMG LEs from normal and ACL-injured subjects during walking.

Conclusions:

• The developed DFT-based pattern analysis technique effectively characterizes EMG LEs during locomotion.
• Phase spectral information is critical for understanding EMG signal dynamics in gait.
• This method provides a robust approach for differentiating between normal and pathological gait patterns based on EMG data.