Human Body Parts Tracking Using Torso Tracking: Applications to Activity Recognition

Abstract

This paper proposed an approach for human body part tracking which is based on torso tracking. The main goal in this paper is tracking main human body parts such as torso, head and hands. In the proposed method for human body part tracking, we are using connected components to improve the detected silhouette (i.e. detected foreground using background subtraction) in order to detect the body parts with respect to torso location and size. We are also using a blob tracking module which is composed of foreground detection, blob detection and blob tracking in order to find the approximate location and size of torso in each frame (i.e. torso tracking). By tracking torso, we will be able to track other body parts based on their location with respect to the torso. Having found torso size and location, the region of certain body parts on the silhouette will be modeled by an Gaussian ellipse using Gaussian blob modeling in each frame with different color in order to show its location, size and pose. The proposed HBPT approach helps us robustly identify and track human body parts in real-time for further use in Human Activity Recognition in vision-based HCI.