Upper Limb End-Effector Force Estimation During Multi-Muscle Isometric Contraction Tasks Using HD-sEMG and Deep...

Area of Science:

• Biomechanics
• Neuroscience
• Robotics

Background:

• Accurate end-effector force estimation is crucial for controlling robotic systems and understanding human movement.
• Multi-muscle contractions present challenges in isolating the contribution of individual muscles to the overall force output.

Purpose of the Study:

• To determine if individual muscles or combined muscle activation is more effective for end-effector force estimation.
• To develop a model for predicting force based on high-density surface electromyography (HD-sEMG) signals.

Main Methods:

• Collected HD-sEMG data from upper arm and forearm muscles during elbow flexion and palm-pressing tasks.
• Utilized Principal Component Analysis (PCA) to extract representative muscle signals.
• Employed a Deep Belief Network (DBN) to model the relationship between HD-sEMG signals and measured forces.

Main Results:

• Dominant muscles with higher activation levels were found to be more effective in tracking end-effector force variations.
• Individual dominant muscles provided more accurate force estimation compared to combinations of muscles.
• The DBN model achieved high accuracy in force estimation.

Conclusions:

• Individual dominant muscles are superior to combined muscle activation for end-effector force estimation in multi-muscle isometric contractions.
• The proposed Mean Impact Value (MIV) index effectively prioritizes muscles for force estimation based on DBN model performance.