A Prosthetic Hand System by Contralateral-Collaborative Control Based on Multi-task Learning.

Area of Science:

• Biomedical Engineering
• Robotics
• Human-Computer Interaction

Background:

• Conventional prosthetic systems limit bimanual coordination and daily activity capabilities for upper-limb amputees.
• Existing prosthetics lack flexibility and adaptability, impacting user autonomy and quality of life.

Purpose of the Study:

• To develop a novel contralateral cooperative control system for prosthetic hands.
• To enhance prosthetic operational flexibility, grasping performance, and user quality of life.

Main Methods:

• Utilized surface electromyography (sEMG) signals from the intact limb to control the prosthetic hand.
• Developed a hybrid multi-task learning model combining CNN and Transformer for gesture and force recognition.
• Employed bioinspired design, drawing from chameleon tail mechanisms for prosthetic finger enhancement.

Main Results:

• Achieved 96.12% offline accuracy for gesture recognition and 94.57% for force intensity.
• Demonstrated a 93.34% online control success rate with minimal performance gap between amputees and healthy individuals.
• Bioinspired prosthetic fingers showed enhanced anti-torsional stability, anti-slip grip, adaptability, and dexterity.

Conclusions:

• The integrated system enhances user autonomy and quality of life through natural, flexible, and coordinated prosthetic operation.
• The contralateral cooperative control and bioinspired design offer an innovative solution for advanced upper-limb prosthetics.
• The system demonstrates robustness and consistency across diverse users and tasks.