Prediction of Joint Angles Based on Human Lower Limb Surface Electromyography

Area of Science:

• Biomedical Engineering
• Rehabilitation Technology
• Human-Machine Interface

Background:

• Wearable exoskeletons enhance mobility for individuals with impairments.
• Electromyography (EMG) signals precede movement, offering potential for predicting user intention.
• Accurate prediction of joint angles is crucial for effective exoskeleton control.

Purpose of the Study:

• To develop and evaluate a predictive model for lower limb joint angles using surface electromyography (sEMG) signals.
• To compare the performance of a Cuckoo Search optimized Random Forest (CS-RF) algorithm against other machine learning models for this prediction task.
• To assess the model's effectiveness across different locomotion activities: walking, ascending stairs, and ascending inclines.

Main Methods:

• Identified key lower limb muscle sites using OpenSim software.
• Collected sEMG and inertial data during various locomotion tasks.
• Applied a Complete Ensemble Empirical Mode Decomposition with Adaptive Noise (CEEMDAN) algorithm for sEMG noise reduction.
• Extracted time-domain features from processed sEMG signals.
• Calculated joint angles using quaternions and coordinate transformations.
• Developed a prediction model using a Cuckoo Search optimized Random Forest (CS-RF) regression algorithm.

Main Results:

• The CS-RF model demonstrated superior prediction performance for knee and hip joint angles compared to standard Random Forest (RF), Support Vector Machine (SVM), and Back Propagation (BP) neural networks.
• Optimal evaluation metrics (RMSE, MAE, R2) were achieved by the CS-RF model across walking, stair climbing, and uphill walking scenarios.
• The model effectively translated sEMG signals into accurate joint angle predictions.

Conclusions:

• The CS-RF algorithm provides a highly accurate and robust method for predicting lower limb joint angles from sEMG signals.
• This approach holds significant potential for improving the control and responsiveness of wearable exoskeletons in rehabilitation.
• The findings suggest a promising pathway for enhancing human-exoskeleton interaction through advanced signal processing and machine learning techniques.