Human Action Analysis Using Virtual Grounding Point and Motion History

In this paper, we propose an approach to human action analysis for home care monitoring system in the aspect of image processing and life science technologies. We introduce a new concept of a virtual grounding point representing the position of a target person as an innovated feature for action analysis. Specifically, in developing action analysis, the background subtraction is firstly conducted by applying the Mixture of Gaussian and low rank subspace learning. After that, the graph cut is embedded to enhance the foregrounds in order to detect both of moving and motionless object. Secondly, the virtual grounding point is calculated by using the centroid of silhouette image. Finally, motion of the person is estimated by using timed motion history image in order to improve the accuracy of action analysis. A series of the experiments are conducted to confirm the effectiveness of the proposed method.