Concurrent and Continuous Prediction of Finger Kinetics and Kinematics via Motoneuron Activities

Area of Science:

• Neuroscience
• Robotics
• Biomedical Engineering

Background:

• Robust neural decoding is essential for intuitive control of assistive devices like robotic hands.
• Existing decoders often struggle to predict kinetic and kinematic variables concurrently.
• Accurate prediction of finger forces and joint angles is crucial for dexterous hand movements.

Purpose of the Study:

• To develop a continuous neural decoding approach for simultaneous prediction of fingertip forces and joint angles.
• To enable real-time, intuitive control of multi-finger robotic systems.
• To improve the performance of assistive robotic hands for daily tasks.

Main Methods:

• Obtained motoneuron firing activities by decomposing high-density electromyogram (HD EMG) signals.
• Grouped and refined motoneuron activity specific to each finger and task (force/movement).
• Applied refined motoneuron groups to new EMG data in real-time, comparing with EMG-amplitude prediction.

Main Results:

• The developed decoding approach significantly outperformed EMG-amplitude prediction.
• Achieved lower prediction error for both force (3.47±0.43% MVC vs 6.64±0.69% MVC) and joint angles (5.40±0.50° vs 12.8±0.65°).
• Demonstrated higher correlation between estimated and recorded motor output for both force (0.75±0.02 vs 0.66±0.05) and joint angles (0.94±0.01 vs 0.5±0.05).

Conclusions:

• The developed neural decoding algorithm accurately predicts finger forces and joint angles concurrently in real-time.
• This method enhances the potential for intuitive control of assistive robotic hands.
• Enables dexterous hand skills involving both force and dynamic movement control.