Improving regression-based myoelectric control: User compliance, simultaneity, and incremental learning
Loading...
Date
2025-03
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
University of New Brunswick
Abstract
Regression-based myoelectric control offers the potential for simultaneous, independent, proportional control, but current implementations are limited by inconsistent training protocols and robustness issues. This thesis aims to address these limitations by providing a novel training protocol that overcomes key robustness issues that have hindered regression-based myoelectric controllers for decades. Two studies are performed, one demonstrating the volatility of current training protocols and the other proposing an alternative paradigm leveraging context-informed incremental learning. Results first show that models trained using existing protocols are affected by simply changing the visual prompting style. In response, the novel training paradigm is shown to significantly improve performance compared to traditional approaches. Furthermore, multiple co-adaptive approaches are contrasted to demonstrate the importance of building tolerance to user behaviours into myoelectric controllers. The results presented in this work provide important considerations for training myoelectric controllers and emphasize the value of designing with the user in mind.