Carlos Henrique Barbosa da Cunha,  Byron Leite Dantas Bezerra. 

This paper presents a module developed for SudaRAframework focused on hand gestures for interaction inAugmented Reality applications. The technique applied is basedon extracting the contour of the hand pixels, which are convertedinto a frequency distribution of the types of neighborhoodsfound in tracing this contour. The vector resulted from thisprocess is passed to an Artificial Neural Network for trainingand evaluation, mapping this vector to their respective gesture.Thus, the user can register his own gestures through images andthen detect them within his application.

http://www.lbd.dcc.ufmg.br/colecoes/wrva/2011/0035.pdf

Processando consulta.