Reconhecimento de padrões em movimentos de extensão dos dedos da mão utilizando dados coletados por redes de bragg em fibras óticas

Abstract

Research and development of optical sensors have gained significant attention in recent years. The ability to multiplex sensors within a single optical fiber, the reduced volume, and immunity to electromagnetic interference are some factors contributing to the popularization of this technology. In this context, this dissertation aims to apply two FBG sensors as strain gauges to capture signals representing forearm muscle deformation caused by individual extension movements of each finger. These signals can, in the future, be used for the treatment and better understanding of muscle injuries, as well as for the control of intelligent prosthetics. The signals were then processed and analyzed to find features that could distinguish them from each other, using pattern recognition techniques. Four classifiers - Support Vector Machine, Decision Tree, Random Forest, and k-Nearest Neighbors - were trained with twelve hundred finger extension movements, and, in the end, the direct relationship between the classifiers’ accuracy - their combination achieved 92.2.