Real time human motion recognition based on a motion history image and an eigenspace

Abstract

In this paper, we propose an efficient real time motion recognition technique based on the eigenspace method. In the proposed system, we use a motion history image (MHI) (Bobick and Davis, 2001) and a superposed motion image (SMI) to realize real time motion recognition employing an eigenspace. Application fields of the present technique may include a surveillance system tracking a person who behaves in a suspicious manner, or an intelligent rescue robot helping an elder person who is carrying a heavy bag or sitting down on the road as he/she feels bad.