Action interference in simultaneous and proportional myocontrol: comparing force- and electromyography

Abstract

Myocontrol, that is, control of a prosthesis via muscle signals, is still a surprisingly hard problem. Recent research indicates that surface electromyography (sEMG), the traditional technique used to detect a subject's intent, could proficiently be replaced, or conjoined with, other techniques (multi-modal myocontrol), with the aim to improve both on dexterity and reliability. In this paper we present an online assessment of multimodal sEMG and force myography (FMG) targeted at hand and wrist myocontrol. Twenty sEMG and FMG sensors in total were used to enforce simultaneous and proportional control of hand opening/closing, wrist pronation/supination and wrist flexion/extension of 12 intact subjects. We found that FMG yields in general a better performance than sEMG, and that the main drawback of the sEMG array we used is not the inability to perform a desired action, but rather action interference, that is, the undesired concurrent activation of another action. FMG, on the other hand, causes less interference.

Simultaneous and Proportional Real-Time Myocontrol of Up to Three Degrees of Freedom of the Wrist and Hand

“…Furthermore, X stands for the design matrix with all samples of EMG features collected during the algorithm training and Y represents the corresponding hand/wrist configurations. This machine learning (ML) method is one of the fundamental regression techniques, and due to its low computational cost and predictive behaviour it has been used for myoelectric control in previous studies [22], [40]- [42].…”

Use of Advanced Materials and Artificial Intelligence in Electromyography Signal Detection and Interpretation

“…M. Nowak et al [ 211 ] developed a protocol to control the prosthesis, combining EMG with FMG (force myography). They used a data acquisition device, consisting of an ADC and two bracelets with sEMG and FMG sensors to collect these two signals from subjects.…”

Simultaneous Assessment and Training of an Upper-Limb Amputee using Incremental Machine-Learning-based Myocontrol: A Single-Case Experimental Design

“…These technological advancements are accompanied by novel developments in myocontrol, which is the control of (prosthetic) devices using muscle signals, most commonly based on electromyography (EMG). These developments are manifold and include the dis-tinction of up to 11 intended actions, such as power grasp, pointing index or wrist flexion, with a success rate above 94% [2], the usage of high-density sensor matrices for control of up to 4 degrees of freedom (DOFs) of a robotic arm [3] or for decoding spike trains [4,5], feature extraction based on deep learning [6] or the usage of different sensor modalities, such as forcemyography [7,8,9,10], ultrasound [11,12] or electrical impedance tomography [13,14]. Yet, the clinical standard since decades is a two-electrode control that uses a switching command, where a trigger is used where the EMG has to meet specific demands, to cycle through the DOFs of the prosthetic setup [15].…”

“…An early involvement of the user is essential, since findings offline (without the user in the loop) do not translate to the online application (with the user in the loop) [18]. Hence, online testing with the user performing goal-reaching tasks has become the standard in evaluating the performance of a novel method [8,19,20,21,22]. Moreover, the introduction of ML-based methods adds a further processing step to myocontrol.…”

“…In this work we performed a long-term study involving one transradial amputee fitted with a multiarticulated prosthetic hand and a custom-built socket controlled by the incremental myocontroller based on Ridge Regression with Random Fourier Features (RR-RFF) [7,8,25,32,33]. Using the SATMC procedure allowed the participant to train how to use the incremental myocontroller, while simultaneously assessing the performance of the user in daily-living tasks.…”