Ground Reaction Force and Moment Estimation through EMG Sensing Using Long Short-Term Memory Network during Posture...

Area of Science:

• Biomechanics
• Robotics
• Machine Learning

Background:

• Human motion prediction traditionally relies on kinematic data.
• Estimating dynamic information like ground reaction force (GRF) is crucial for advanced prediction.
• Forces are the origin of motion, making dynamic analysis essential.

Purpose of the Study:

• To propose a novel method for estimating GRF and ground reaction moment (GRM) using electromyography (EMG) and inertial measurement unit (IMU) sensors.
• To leverage machine learning, specifically a long short-term memory (LSTM) network, for accurate motion dynamics prediction.
• To develop a visualization system for GRF in 3D space to predict motion direction.

Main Methods:

• Utilized a long short-term memory (LSTM) network trained on EMG and IMU data.
• Applied the method to estimate GRF during posture control and stepping motions.
• Developed a GRF visualization system within a Unity environment for real-time motion prediction.

Main Results:

• Achieved GRF estimation with a root mean square error (RMSE) of 8.22 ± 0.97% for posture control and 11.17 ± 2.16% for stepping motion.
• Confirmed the essential role of EMG input for predicting GRF and GRM, especially with limited sensors.
• Successfully visualized GRF vectors in 3D space and predicted motion direction.

Conclusions:

• The proposed machine learning approach effectively estimates GRF and GRM from EMG and IMU data.
• EMG is vital for accurate dynamic motion prediction, particularly in sensor-limited scenarios.
• The GRF visualization system offers valuable insights for human motion prediction using portable sensors.