Electromyography (EMG) Signal Contributions in Speed and Slope Estimation Using Robotic Exoskeletons.

Area of Science:

• Robotics
• Biomedical Engineering
• Machine Learning

Background:

• Robotic exoskeletons can enhance community ambulation for aging populations.
• Current exoskeleton controllers rely on environmental data like speed and slope, but struggle with dynamic changes.
• Machine learning models offer a potential solution by integrating user's biological signals and mechanical sensor data.

Purpose of the Study:

• To develop a neural network-based model for estimating walking speed and slope for a powered hip exoskeleton.
• To investigate the contribution of electromyography (EMG) signals to the model's performance in static and dynamic conditions.
• To analyze the impact of different EMG electrode placements on model accuracy.

Main Methods:

• A neural network model was developed to estimate walking speed and slope using mechanical sensor data and electromyography (EMG) signals.
• The model's performance was evaluated with and without EMG data in both static and dynamic walking scenarios.
• Electrode placement variations were tested to determine optimal configurations for EMG signal acquisition.

Main Results:

• The machine learning model achieved low error rates: below 0.08 m/s RMSE for speed and 1.3 RMSE for slope.
• Incorporating EMG signals reduced the error rate by 14.8% compared to using only mechanical sensors.
• Optimal performance was achieved when EMG electrodes were placed within the exoskeleton's interface region.

Conclusions:

• EMG signals significantly enhance the accuracy of walking speed and slope estimation in powered hip exoskeletons.
• The developed machine learning model improves robotic assistance for community ambulation in aging individuals.
• Strategic EMG electrode placement is crucial for maximizing the benefits of EMG-integrated exoskeleton control.