Tabletop Human-Object Interaction using Depth Layering

Abstract

In this paper, we propose an efficient tabletop human-object interaction system by categorizing depth information. This system uses three layers to apply proper algorithms respectively. Background layer is the trained background environment model which is initialized at the beginning. The static layer is used to store steady objects which is registered for interaction subject. And the dynamic layer handles moving objects which in this case is hand. In each layer, a proper algorithm can be applied separately without interference. The system will show how effective this approach works.