Dynamic posture estimation in a network of depth sensors using sample points
S. D. Maryam Rasouli, Shahram Payandeh · 2017
In this paper, we propose a novel method to estimate the posture of the human body in a network of depth sensors. We divide the body into two regions and take a limited number of selective samples from point cloud of human body. The sample points are utilized to compute 2D gradient and the histograms of the gradient of sample points along two different scan directions. These extracted values are then selected as feature points of each posture which are then compared with stored values for further posture classification. The input depth map is assigned to a known posture having the most similarity among all available samples. It is shown that the proposed approach offers high recognition rate despite low number of training set.