A novel approach to improve background subtraction method for fall detection system

The fall detection system composed in general from three models, the first is a detection moving object, the second is tracking moving object and the third is recognition person's behavior. In this paper, we will deal out the first module which is the main unit in the fall detection system, because it is where we detect the person in the image. There are a lot of approaches in this way, but their results are not perfect due to the change of the background model. Many works propose to improve those results, but these approaches are not satisfied in some states. In this way, we propose a novel approach by using the optical flow which can improve The result to extract the foreground human body without errors. © 2015 IEEE.