Context-Informed Incremental Learning Improves Throughput and Reduces Drift in Regression-Based Myoelectric Control

Abstract

Despite decades of research, commercially available powered myoelectric prostheses continue to use sequential, classification-based control. While regression-based approaches can improve the dexterity offered through simultaneous, independent, and proportional control, current training protocols lack consistency across studies and fail to capture realistic user behaviours, resulting in robustness issues. To address these challenges, this work employs context-informed incremental learning (CIIL) in an unconstrained, velocity-based environment for regression-based myoelectric control. Two new adaptive models, one inspired by previous works (O-CIIL) and one modified to factor in user compliance and behaviours (T-CIIL), were compared with two models trained using traditional screen-guided training. Sixteen participants completed an online Fitts’ Law target acquisition task. Both adaptive approaches significantly outperformed ( ${p}\lt {0}.{05}$ ) the non-adaptive models across a variety of metrics. Additionally, T-CIIL outperformed O-CIIL in alleviating drift and action interference, key issues that have plagued existing regression-based myoelectric control systems. These findings are supported by two novel metrics, namely action interference and simultaneity gain, which show that adding simultaneity often increases instability in the form of undesired and uncontrollable simultaneous motions. These findings demonstrate the viability of CIIL in an unconstrained, velocity-controlled environment for regression-based myoelectric control, and highlight the importance of capturing user behaviours when training regression-based myoelectric control systems. Source code is available on https://github.com/cbmorrell/adaptive-regression