A 3-DOF hemi-constrained wrist motion/force detection device for deploying simultaneous myoelectric control.

Area of Science:

• Biomechanics
• Neuroscience
• Robotics

Background:

• Simultaneous electromyography (EMG) control for wrist prosthetics often uses force or movement as training targets.
• The complex relationship between wrist force and movement limits the precision of single-target control.
• Existing methods lack a comprehensive approach to capture the full dynamics of wrist state.

Purpose of the Study:

• To introduce a novel platform for synchronous acquisition of three degrees of freedom (DOF) wrist motion/force with multi-channel EMG signals.
• To establish a stable, direct mapping between wrist movement and force using a custom-built device.
• To enable precise prediction of arbitrary 3-DOF wrist movements using machine learning models.

Main Methods:

• Development of a hemi-constrained platform for simultaneous 3-DOF wrist motion/force and EMG acquisition.
• Implementation of a self-made wrist force-movement mapping device for stable signal correlation.
• Utilizing a laser cursor for direct, real-time feedback of 3-DOF wrist movement without decoupling algorithms.
• Employing a support vector regression model trained on acquired data for movement prediction.

Main Results:

• The platform successfully acquired synchronous multi-channel EMG and 3-DOF wrist motion/force data.
• The support vector regression model demonstrated high accuracy in predicting arbitrary 3-DOF wrist movements.
• Cross-validation showed that the regression accuracy for free 3-DOF movements was comparable to that of regular 2-DOF movements (p > 0.1).

Conclusions:

• The proposed platform provides a precise and comprehensive method for capturing wrist dynamics.
• Accurate prediction of 3-DOF wrist movements is achievable, paving the way for advanced prosthetic control.
• This system overcomes limitations of single-target control by integrating force, movement, and EMG data.