Progressive unsupervised control of myoelectric upper limbs.

Area of Science:

• Biomedical Engineering
• Rehabilitation Robotics
• Human-Computer Interaction

Background:

• Unsupervised myocontrol methods for prostheses avoid complex training data but often predefine function numbers.
• Existing methods limit users by not accommodating evolving motor skills or new function discovery.
• This gap hinders natural adaptation and skill development in myoelectric prosthetic users.

Purpose of the Study:

• To introduce and evaluate a progressive unsupervised myocontrol (PUM) paradigm.
• To enable users and control models to coadaptively identify muscle synergies for prosthetic control.
• To allow users to progressively learn new prosthetic functions as their skills develop.

Main Methods:

• Developed the PUM paradigm for coadaptive learning of muscle synergies and prosthetic functions.
• Users started with one function and could request more upon mastery.
• Conducted a multi-session user study comparing PUM against a non-progressive baseline using a virtual hand and wrist.
• Included participants with upper-limb differences and non-disabled controls.

Main Results:

• PUM facilitated autonomous learning of 3-4 myocontrol functions in both participant groups.
• Participants with limb differences showed comparable success rates but faced challenges in skill internalization and movement quality.
• Performance with PUM was comparable to the non-progressive baseline in non-disabled participants.
• Both existing and newly identified muscle synergies were utilized for control.

Conclusions:

• The PUM paradigm supports autonomous learning and adaptation to users' motor skills.
• It accommodates varied preexisting motor skills and facilitates the development of new skills over time.
• PUM offers a flexible approach to myoelectric control, enhancing user independence and prosthetic functionality.