Sistema para rastreamento de gestos para realidade virtual

Abstract

This work presents an improvement to an open-source virtual reality headset. One of the barriers to total imersion and simulation for virtual reality headsets are the controllers, which are not capable of reproducing the high level of detail of the human hand movement. The project idea is to provide to the virtual reality headset users a low cost way to interact with the virtual environment using only their hands. To achieve this goal, a prototype of a computer vision system using a webcam was built to perform hand-pose estimation on the user hands and transmit this estimation to a virtual environment developed in Unity. As the result, a comparison of the developed model with the Leapmotion commercial hand recognition system was presented, where advantages regarding the range of motion and interaction between the hands and disadvantages in the performance of image capture and processing were identified and discussed, being more evident in quick movements.