(Un)supervised (Co)adaptation via Incremental Learning for Myoelectric Control: Motivation, Review, and Future...

Area of Science:

• Biomedical Engineering
• Machine Learning
• Rehabilitation Technology

Background:

• Traditional myoelectric control faces challenges with signal non-stationarities and requires frequent recalibration.
• Incremental learning presents a potential solution for adaptive prosthetic systems by enabling continuous model updates.
• Existing methods struggle with user-specific variations and environmental changes.

Purpose of the Study:

• To provide a narrative review of incremental learning methods for myoelectric control.
• To outline the historical trajectory and future potential of adaptive prosthetic systems.
• To establish a framework for evaluating current research and guiding future innovations in the field.

Main Methods:

• A taxonomy of incremental learning strategies is presented, categorizing methods into four types: dedicated on-demand recalibration, unsupervised, predictor-dependent, and environment-dependent incremental learning.
• The review discusses the methodology, strengths, and limitations of each category.
• Three settings for incremental learning (domain-incremental, task-incremental, class-incremental continual learning) are established.

Main Results:

• The paper categorizes incremental learning strategies and settings, providing a framework for analysis.
• Emerging trends like transfer learning, domain adaptation, and self-supervised regression are highlighted.
• Potential advancements include physiologically-inspired algorithms, novel end-effector designs, and human-device co-adaptation.

Conclusions:

• Incremental learning offers a paradigm shift for myoelectric control, enabling continuous adaptation and addressing limitations of traditional methods.
• Open challenges include distinguishing signal changes, balancing model complexity, and managing user-model adaptation.
• These insights are crucial for developing next-generation myoelectric systems that are robust, intuitive, and adaptable.