Sistema web com estimativa de pose e análise de movimento para aplicações de auxílio à prática de exercício físico

Abstract

The regular practice of physical exercises provides benefits to physical and mental health, as well as controlling or mitigating diseases. It is recommended that exercises are performed under the supervision of a physical education professional, because if they are not, there is a great risk of injury or lack of effective muscle activation. The technology for tracking parts of the human body, such as Google's BlazePose, makes it possible to identify physical activities, but there is currently no software available that is independent of the exercise to provide data on range of motion, execution time and training frequency. The present monograph aims to build a software system that can provide this data, returning a more complete evaluation of the exercise done to the instructor and student. Such a system takes advantage of features from human pose tracking, BlazePose, and also from software libraries such as Flask and React. The system design applied software engineering tools like use case diagram, class diagram; Mathematical modeling of the data obtained is also carried out to find joint angles and full motion detection. The results showed that the system has the most important features and that it is possible for the instructor to evaluate student data to find those who need help. Thus, the instructor's monitoring can be done in a more indirect way, without being presently in the execution of the exercise, and with more accuracy.