C0NTR0LL: sistema de reconhecimento de gestos e gerenciamento de ações em jogos digitais

Abstract

In the last decade, as machine learning techniques and input devices for capturing and estimating gestures and poses have evolved significantly, providing a new form of interaction between man-machines and promoting virtual environments as a recovery treatment through telerehabilitation. In order to assess what is available for the context of telerehabilitation with digital games, a systematic review is presented with the main control systems and devices in the research area. Despite the many therapeutic advantages and benefits of this new paradigm, it is important to centralize the estimation and classification of the complete pose in a single environment, ensuring a safe, useful, scalable and ergonomic tool. This work proposes to address all these points, through the C0NTR0LL motor telerehabilitation system. System scalability is handled by implementing a modular architecture. The safety and effectiveness of the tool are guaranteed by an artificial intelligence module that assesses the quality of movements created by rehabilitation professionals and performed by the user. The usability of the system is evaluated by a method of content validity and the usability scale of the system. The validation of the questionnaires is evaluated using Cronbach’s Alpha coefficient. The results show that a centralization of poses of the system is able to assess the accuracy of human movement both for model training and for application in rehabilitation preparation.