Design, Control, and Sensory Feedback of Externally Powered Hand Prostheses: A Literature Review

Abstract

In recent years, there has been a steep rise in the quality of prostheses for patients with upper limb amputations. Researchers have begun to identify methods of making prosthetic hands both functional and cosmetically appealing, in contrast to past designs. Many improvements have occurred because of novel design strategies, such as the use of underactuated mechanisms, which allow for more degrees of freedom (DOF) or help reduce the weight of the prosthesis. The increase in functionality is also due in large part to advancements in control strategies for prosthetic hands. One common control method, using electromyographic (EMG) signals generated by muscle contractions, has allowed for an increase in the DOF of hand designs and a larger number of available grip patterns with little added complexity for the wearer. Another recent improvement in prosthetic hand design instead employs electroneurographic (ENG) signals, requiring an interface directly with the peripheral nervous system (PNS) or the central nervous system (CNS). Despite the recent progress in design and control strategies, however, prosthetic hands are still far more limited than the actual human hand. This review outlines the recent progress in the development of electrode-based prosthetic hands, detailing advancements in the areas of design, sensory feedback, and control through EMG and ENG signals (with a particular focus on interfaces with the PNS). The potential benefits and limitations of both control strategies, in terms of signal classification, invasiveness, and sensory feedback, are discussed. Finally, a brief overview of interfaces with the CNS is provided, and potential future developments for prosthetic hand design are discussed.

NLR, MLP, SVM, and LDA: a comparative analysis on EMG data from people with trans-radial amputation

“…Linear classifiers, such as Linear Discriminant Analysis (LDA), Logistic Regression (LR) or Support Vector Machine (SVM) with linear kernel, and nonlinear classifiers, such as Non-linear Logistic Regression (NLR), SVM with nonlinear kernels and Multi-Layer Perceptron (MLP), represent the state-of-the-art about pattern recognition classifiers [ 14 , 15 ]. The main difference between linear and nonlinear classifiers consists in the shape of the decision boundary: straight line, or plane in the first case and curved line, or surface, in the second.…”

Investigation of Channel Selection for Gesture Classification for Prosthesis Control Using Force Myography: A Case Study

“…Various invasive and non-invasive HMIs have been introduced for the control of upper limb prostheses. Some of the non-invasive HMIs for this application that have gained interest in the research community are: gaze tracking (Castellini and Koiva, 2013), electromyography (EMG) (Castellini et al, 2009;Scheme and Englehart, 2011), electroneurography (ENG) (Cloutier and Yang, 2013), mechanomyography (MMG) (Xiloyannis et al, 2015), force myography (FMG) (Rasouli et al, 2015), etc.…”

Tactile sensing and feedback in SEMG hand

“…Recently it was exploited in the Yokoi Hand to deliver grip force feedback on a single skin-site and by Geng et al (2011) to investigate the effects of different patterns on the perception threshold. Electrotactile feedback system is comprised of force sensors that are placed on the fingers and palm of a prosthetic hand, interface circuits for processing the sensor data and electrodes that are placed on nearby skin (Cloutier and Yang, 2013). It communicates sensory information to the prosthetic user via electrodes placed on the user's skin.…”