MMG-Based Knee Dynamic Extension Force Estimation Using Cross-Talk and IGWO-LSTM.

Area of Science:

• Biomedical Engineering
• Neuroscience
• Sports Science

Background:

• Mechanomyography (MMG) measures muscle physiological activity, reflecting motor unit recruitment and contraction frequency.
• MMG can estimate skeletal muscle force but is hindered by cross-talk and time-series correlation, impacting dynamic force estimation accuracy.
• Accurate dynamic muscle force estimation is crucial for developing advanced wearable and assistive devices.

Purpose of the Study:

• To investigate if using MMG signals from a single muscle with reduced cross-talk improves knee dynamic extension force estimation accuracy.
• To develop and validate a novel deep learning model for enhanced dynamic knee extension force estimation.
• To improve the flexibility and interaction capabilities of future rehabilitation and assistive devices.

Main Methods:

• Proposed a hypothesis that single-muscle MMG signals with less cross-talk enhance force estimation accuracy.
• Validated the hypothesis by comparing estimation results from various muscle signal feature combinations.
• Developed an improved grey wolf optimizer optimized long short-term memory network (IGWO-LSTM) for force estimation.

Main Results:

• MMG signals from a single muscle with less cross-talk demonstrated superior performance in estimating dynamic knee extension force.
• The proposed IGWO-LSTM model achieved the best performance metrics compared to existing state-of-the-art models.
• The study confirmed the hypothesis regarding the benefits of reduced cross-talk in MMG signals for force estimation.

Conclusions:

• Utilizing MMG signals from a single muscle with minimized cross-talk is effective for accurate dynamic knee extension force estimation.
• The IGWO-LSTM model represents a significant advancement in MMG-based force estimation, outperforming other models.
• This research contributes to understanding quadriceps contraction mechanisms and advancing wearable rehabilitation and assistive technologies.