Classification of Multiple Finger Motions During Dynamic Upper Limb Movements

Area of Science:

• Biomedical Engineering
• Rehabilitation Robotics
• Human-Computer Interaction

Background:

• Restoring human hand function requires advanced prostheses capable of operating in diverse daily conditions.
• Myoelectric control schemes are sensitive to variations from muscle contractions, arm movements, and external forces.

Purpose of the Study:

• To investigate methods for robust classification in pattern recognition-based myoelectric control.
• To evaluate the impact of different training paradigms on myoelectric signal variations.

Main Methods:

• Examined four distinct training paradigms (data-collection protocols) for myoelectric control.
• Quantified the effectiveness of each paradigm in achieving robust signal classification.
• Analyzed classification accuracy based on arm/wrist motion primitives.

Main Results:

• The dynamic posture and dynamic effort (DPDE) training paradigm significantly mitigated misclassification rates.
• Finger motion misclassification correlated strongly with wrist pronation/supination, less so with arm position.
• Integrating proprioceptive hand orientation data offered only marginal improvements in classification accuracy.

Conclusions:

• The DPDE training paradigm enhances the robustness of myoelectric control for prosthetic hands.
• Understanding the influence of specific movements like wrist pronation is crucial for improving control accuracy.
• Future research may explore combined sensory feedback for more intuitive prosthetic control.