Performance analysis of myoelectric regressors trained with transfer learning

• Grant number: 21/09696-4
• Support Opportunities: Scholarships in Brazil - Scientific Initiation
• Start date: April 01, 2022
• End date: September 30, 2022
• Field of knowledge: Engineering - Biomedical Engineering - Bioengineering
• Principal Investigator: Arturo Forner Cordero
• Grantee: Pedro Pereira Moraes
• Host Institution: Escola Politécnica (EP). Universidade de São Paulo (USP). São Paulo , SP, Brazil

Abstract

Prostheses and exoskeletons capable of extracting movement intentions from the electromyogram (EMG) have the potential to prevent or mitigate motor difficulties. The motion intent extraction would be performed by a myoelectric controller, which consists of a machine learning algorithm that would solve classification or regression problems. However, there are still obstacles to the practical use of this technology. One of the challenges for the development of myoelectric regressors is the insufficiency of online data available for regression. Even though there are several online EMG databases that could be used to improve the performance of myoelectric controllers, most of them are focused on classification. Because of that, this project aims to evaluate whether through transfer learning (TL) it is possible to make use of data aimed at classification to increase the accuracy of regressors. (AU)