RGB-D Hand Pose Estimation Using Fourier Descriptor

Abstract

The emergence of civil-use depth sensors made vision-based man-machine interaction possible through hand gesture, which relies fundamentally on hand pose estimation techniques. In this work, a novel hand pose estimation method is proposed, based essentially on Fourier-descriptor-based indexing in a pre-built hand template library. The proposed method takes color-depth as input, extracts necessary hand information in preprocessing, and estimate hand pose through two stages, from rough to accurate. This proposed hand pose estimation method is simple and efficient; quantitative and qualitative evaluations exhibit its effectiveness.