Effort classification in running using machine learning through ground reaction force and facial recognition

• Grant number: 24/15658-6
• Support Opportunities: Scholarships in Brazil - Doctorate
• Start date: March 01, 2025
• End date: August 31, 2028
• Field of knowledge: Health Sciences - Physical Education
• Principal Investigator: Paulo Roberto Pereira Santiago
• Grantee: Abel Gonçalves Chinaglia
• Host Institution: Escola de Educação Física e Esporte de Ribeirão Preto (EEFERP). Universidade de São Paulo (USP). Ribeirão Preto , SP, Brazil

Abstract

The proposed research aims to develop a method for classifying running patterns associated with perceived effort, utilizing characteristics of facial expressions and ground reaction force signals. The study stands out for its methodological approach, integrating machine learning techniques, facial expression analysis, and ground reaction force data on a low-cost treadmill. The methodology involves collecting data from participants during exertion sessions on a treadmill, with simultaneous recording of subjective effort perception and facial expressions. Ground reaction force signals are captured by sensors integrated into the treadmill, while facial expressions are recorded by a digital camera. The data are pre-processed and fused, allowing the construction of a machine learning model using the Multilayer Perceptron (MLP) algorithm with backpropagation. The research face scientific and technological challenges, such as integrating different types of data, preprocessing and merging information, and implementing an effective machine learning model. It is expected that the results will significantly contribute to the development of more accessible and accurate methods for assessing effort during running, with a potential impact on injury prevention and training enhancement. The study represents an important contribution to the field of human motion analysis, standing out for the integration of different methodological approaches and the application of low-cost technologies.