Development and testing of a virtual simulator for a myoelectric prosthesis prototype – the PRISMA Hand II – to improve its usability and acceptability. (May 2023)

• Record Type: Journal Article
• Authors: Leccia, Adriano; Sallam, Mohamed; Grazioso, Stanislao; Caporaso, Teodorico; Di Gironimo, Giuseppe; Ficuciello, Fanny
• Is Part Of: Engineering applications of artificial intelligence. Volume 121(2023)
• Publication Date: 2023-05
• Subjects: Rehabilitation robotics -- Active prosthetic hand -- sEMG signal processing -- Pattern recognition -- Multi-Layer Perceptron (MLP) -- Biomechatronic application; Engineering -- Data processing -- Periodicals; Artificial intelligence -- Periodicals; Expert systems (Computer science) -- Periodicals; Ingénierie -- Informatique -- Périodiques; Intelligence artificielle -- Périodiques; Systèmes experts (Informatique) -- Périodiques; Artificial intelligence; Engineering -- Data processing; Expert systems (Computer science); Periodicals; 620.00285
• DOI: 10.1016/j.engappai.2023.105853
• Languages: English
• ISSNs: 0952-1976

Abstract

Abstract: Artificial limbs can help people missing body parts to restore some of their daily-life activities. However, the user should spend up to a few months to intuitively control the new device. During this period, she/he may suffer pain due to wearing or using the prosthesis inappropriately. This research presents a virtual simulator that allows the user to carry out training sessions for controlling the prosthesis. A set of Surface Electromyographic (sEMG) sensors are used to acquire the signals from user's muscles and send them to a recognition algorithm that interprets the patient's intentions. Simultaneously, the patient observes the response of her/his device on the simulator. Two studies are presented: the first study evaluate the performance of three different recognition algorithms i.e., Linear Discriminant Analysis (LDA), Support Vector Machine (SVM), and Multi-Layer Perceptron (MLP), based on the successful recognition of the patient's intentions. The second study investigates the least number of sEMG sensors to be used, as having less components improves the patient's wearability and decreases the processing time. The developed simulator represents a real prosthetic device, PRISMA hand II. The results showed the superiority of the MLP with 80% of successful recognition when 6-sEMG sensors are used. If a reduced set of gestures is considered (frequently needed by the patient), 90% of successful recognition could be achieved. Less sEMG sensors significantly degraded the performance of the recognition algorithm as only 53.8% of successful recognition could be achieved. All experiments were conducted with the help of a patient with below-elbow amputation.