Directional Forgetting for Stable Co-Adaptation in Myoelectric Control

Abstract

Conventional myoelectric controllers provide a mapping between electromyographic signals and prosthetic functions. However, due to a number of instabilities continuously challenging this process, an initial mapping may require an extended calibration phase with long periods of user-training in order to ensure satisfactory performance. Recently, studies on co-adaptation have highlighted the benefits of concurrent user learning and machine adaptation where systems can cope with deficiencies in the initial model by learning from newly acquired data. However, the success remains highly dependent on careful weighting of these new data. In this study, we proposed a function driven directional forgetting approach to the recursive least-squares algorithm as opposed to the classic exponential forgetting scheme. By only discounting past information in the same direction of the new data, local corrections to the mapping would induce less distortion to other regions. To validate the approach, subjects performed a set of real-time myoelectric tasks over a range of forgetting factors. Results show that directional forgetting with a forgetting factor of 0.995 outperformed exponential forgetting as well as unassisted user learning. Moreover, myoelectric control remained stable after adaptation with directional forgetting over a range of forgetting factors. These results indicate that a directional approach to discounting past training data can improve performance and alleviate sensitivities to parameter selection in recursive adaptation algorithms.

Assistance Robotics and Biosensors 2019

“…Other issues that affect classification accuracy are evaluated in [2,3]. An adaptive classification model based on directional forgetting is proposed in [2].…”

“…Other issues that affect classification accuracy are evaluated in [2,3]. An adaptive classification model based on directional forgetting is proposed in [2]. This novel algorithm addresses signal instability issues through a calibration of the model in time, showing good results in a small number of volunteers.…”

A Framework for Optimizing Co-adaptation in Body-Machine Interfaces

“…However, extensive usage requires interface stability, which is at present a considerable challenge both due to technological characteristics of the device, and due to physiological and functional processes active at the user's level (Young et al, 2011;Barrese et al, 2013;Orsborn et al, 2014;Downey et al, 2018). Co-adaptive algorithms for HMIs have been developed to address the issue of decoder instability (Vidaurre et al, 2011;Kao et al, 2017;Yeung et al, 2019;Degenhart et al, 2020;Silversmith et al, 2020) and to compensate for performance degradation due to the emergent closed-loop dynamics during use (Orsborn et al, 2012;Dangi et al, 2013;Shenoy and Carmena, 2014;Hahne et al, 2015;De Santis et al, 2018). One goal of these strategies is to reduce reliance on user adaptation to compensate for imperfections in the interface, a process that can be lengthy and cognitively demanding, besides being often insufficient for guaranteeing efficient control (Sadtler et al, 2014;Golub et al, 2018).…”

Co-adaptive myoelectric control for upper limb prostheses

“…Increasing the amount of training data helps to deal with unknown inputs. Yeung et al (2019) proposed a new re-training system, in which depending on the new data added to the training set specific old data was erased. This directional forgetting deleted old data that was in the same direction that the new one added.…”

“…Expanding the training set proved to have benefits against shorter data sets, so the idea was to increase it with the test data and re-calibrate the system with the expanded training set. Yeung et al (2019) proposed a similar method but with a directional paradigm. As newer data was being added to the training set, and older data with similar information was being erased.…”