Desenvolvimento de protótipo para identificação de movimentos de boxe

Abstract

Due the lack of devices that are capable of assist athletes on field during the practice of sports, it was decided to develop a prototype capable of count and identify the main boxing movements. The research presents a theoretical and empirical approach of inertial sensors, neural network knowledge, FIR base and basic boxing knowledge. The main goal of the project is to use inertial sensors inside boxing gloves to identify the types of punches that are being thrown by the athlete. During the project its obstacles and needs were discussed, like the sensor requirements, which punches to identify and the structure of the neural network. Since the chosen device, with inertial sensors, has Bluetooth connectivity, it was developed an Android application to perform the data acquisition of ideal movements and perform the simulation of the trained neural network . The data acquisition of the movements, those being the jab, cross, uppercut and hook, was performed using two professional boxing athletes with 14 years of experience, providing credibility to the data. At the end, the project brings results of the test sessions performed, presenting an 84,8% average rate of success on the identification of the movements.