Beyond the Legs: Leveraging Trunk and Driving Muscle EMG to Improve Gait Model Predictions

Area of Science:

• Biomechanics
• Robotics
• Neuroscience

Background:

• Electromyography (EMG) signals from lower-limb muscles are traditionally used for gait prediction in exoskeletons.
• Trunk muscles play a crucial role in clinical gait analysis, suggesting their potential utility in predictive models.
• Identifying key muscle groups, termed "driving" muscles, may optimize EMG input for enhanced kinematic predictions.

Purpose of the Study:

• To evaluate the predictive accuracy of gait models incorporating trunk muscle EMG versus solely lower-limb EMG.
• To assess the impact of focusing on "driving" muscles for improved gait kinematics prediction.
• To investigate prediction performance across varying time horizons (0.05 ms, 100 ms, 200 ms).

Main Methods:

• Developed gait prediction models using EMG signals from trunk, driving, and lower-limb muscles.
• Predicted hip, knee, and ankle joint angles, and gait cycle percentage.
• Compared model performance based on different muscle signal inputs and prediction horizons.

Main Results:

• Models utilizing trunk muscle EMG significantly outperformed those using only lower-limb EMG, especially for knee angle prediction at longer horizons (p<0.05).
• Incorporating "driving" muscles enhanced prediction accuracy for knee and ankle kinematics, particularly at short prediction horizons (p<0.001).
• Trunk muscle inclusion demonstrably improves gait model accuracy, while "drivers" offer a targeted approach for relevant EMG selection.

Conclusions:

• Trunk muscle EMG signals are valuable additions for enhancing the accuracy of lower-limb exoskeleton gait prediction models.
• Targeting "driving" muscles provides a more effective strategy for selecting relevant EMG inputs to improve kinematic predictions.
• These findings support the integration of trunk and "driving" muscle data for more robust and responsive exoskeleton control.