An evaluation of RGB-D skeleton tracking for use in large vocabulary complex gesture recognition

Abstract

An essential component of any hand gesture recognition system is the hand detector and tracker. While a system with a small vocabulary of sufficiently dissimilar gestures may work well with approximate estimations of hand locations, more accurate hand position information is needed for the best results with a large vocabulary of complex two-handed gestures, such as those found in sign languages. In this paper we assess the feasibility of using a popular commercial skeleton tracking software solution in a large vocabulary gesture recognition system using an RGB-D gesture dataset. We also provide a discussion of where improvements in existing methods utilizing the advantages of depth-sensing technology can be made in order to achieve the best possible results in complex gesture recognition.