EMG-based Simultaneous Estimations of Joint Angle and Torque during Hand Interactions with Environments

Area of Science:

• Neuroscience
• Robotics
• Biomechanics

Background:

• Controlling contact forces during object manipulation requires modulating joint stiffness via agonist-antagonist muscle co-contraction, in addition to position control.
• Accurate real-time estimation of joint kinematics and kinetics is crucial for understanding and replicating human motor control.
• Electromyography (EMG) signals from muscle pairs offer a potential pathway for inferring neural control signals related to limb movement.

Purpose of the Study:

• To develop and validate a novel decoding technique for simultaneously estimating limb joint position and torque in real time.
• To utilize electromyography (EMG) data from agonist-antagonist muscle pairs as input for the decoding algorithm.
• To assess the performance of long short-term memory (LSTM) networks in decoding joint kinematics and kinetics during human-environment interaction.

Main Methods:

• A real-time decoding technique was developed using electromyography (EMG) signals from agonist-antagonist muscle pairs.
• Long short-term memory (LSTM) networks, both unidirectional and bidirectional, were employed as the core processing units.
• Validation involved a robotic setup where the wrist joint moved along a trajectory against controlled resistance.

Main Results:

• The decoding approach achieved over 93% agreement for kinetics (torque) estimation between actual and estimated values.
• The technique demonstrated over 83% agreement for kinematics (angle) estimation during interactions with the environment.
• No significant performance difference was observed between unidirectional and bidirectional LSTM networks for this decoding task.

Conclusions:

• The developed EMG-based decoding technique accurately estimates joint position and torque in real time during dynamic interactions.
• LSTM networks are effective for learning complex time-series relationships in muscle activity for motor decoding.
• This method holds promise for applications in prosthetics, robotics, and neurorehabilitation by providing insights into motor control.