Software that seizes through Kinect, the data of manual systems and it translates to text

Abstract

Proceed. are presented the results of the Project that has as a purpose of developing a Software, that allows, through the depth camera of Kinect. seize the information of the manual signals, that a person carries out in front of Kinect, and shows the text equivalent of this sign. The software presented it is still en in the stage of the prototype are various the improvements that we should and can implement with the purpose of obtaining better results in signs recognition.

The prototype developed it is able for taking the scene points captured through the depth Device of Kinect, apply a screening for delete unnecessary piece of information and after, shows the mesh (something that rebuilds the picture of the final scene in 2D), establishing the difíerence of distances with the color change. This detects the location of the hand or hands in Mesh, then takes pictures of manual signáis and compares it with the pictures stored in the file, these have the respective translation if it finds a picture with a higher level of a coincidence it returns with the correct text. The results of the investigation allow recognizing the stages, and ideal platforms processing the implemented pictures in the prototype.