Current Proceedings on Technology
Yazarlar: Tudor Ioan Cerlinca
Konular:-
Anahtar Kelimeler:Hand gesture recognition,Depth map,Dynamic Time Warping
Özet: The aim of this paper is to present a fast and robust algorithm for the real-time recognition of 3D hand gestures. The algorithm that we propose is based on an improved version (in both speed and accuracy) of the classical Dynamic Time Warping. The algorithm takes as input the depth maps acquired from a stereo-vision camera system (thus, being less sensitive to change of lighting) and it is capable of recognizing 15 different gestures (from a distance ranging from 1 to 5 meters) performed with one or even both hands with an overall accuracy of about 95%. The experimental results showed that the algorithm is suitable for the integration within various Human-Computer/Robot-Interaction-based systems.