Towards a Physiologically Informed Machine-Learning Approach to Estimate Hand Kinematics in a Human-Machine Interface Based on High-Density Myoelectric Signals

• Grant number: 25/03022-2
• Support Opportunities: Scholarships in Brazil - Doctorate
• Start date: July 01, 2025
• End date: June 30, 2028
• Field of knowledge: Engineering - Biomedical Engineering - Bioengineering
• Principal Investigator: Leonardo Abdala Elias
• Grantee: Cristian David Guerrero Mendez
• Host Institution: Faculdade de Engenharia Elétrica e de Computação (FEEC). Universidade Estadual de Campinas (UNICAMP). Campinas , SP, Brazil
• Company: Universidade Estadual de Campinas (UNICAMP). Faculdade de Engenharia Elétrica e de Computação (FEEC)
• Associated research grant: 20/09838-0 - BI0S - Brazilian Institute of Data Science, AP.PCPE

Abstract

Upper limb movements are essential for daily living activities, which are fundamental to maintaining life, such as eating and manipulating objects. The development and implementation of human-machine interfaces have enabled significant advances in rehabilitation systems, facilitating the recovery of these movements in people with severe motor impairments. Data-driven methods have been used to identify or estimate the movement intention of users who employ neural interfaces. However, despite advances in usability in interface control, these methods preclude a complete understanding of the physiology of the motor system during the use of these technologies. This limits the implementation of personalized rehabilitation programs, the understanding and monitoring of the neuromuscular system during interventions, the implementation of more physiological neurofeedback, and the identification of potential biomarkers associated with neurorehabilitation. Therefore, this doctoral project aims to develop and evaluate a computational framework that combines data-driven methods and biophysical modeling of neuromuscular elements, which allows for the development of neuromorphic computing. The computational approach will be used to estimate the joint kinematics of the fingers using high-density electromyography signals. By integrating physiologically based models, the aim is to improve the kinematic estimation of continuous variables, thereby improving the understanding of neuromuscular function and dysfunction and the feasibility of more personalized neurorehabilitation systems. This research could pave the way for the next generation of physiologically informed machine-learning methods and personalized human-machine interfaces.