Identification of People by Measuring Body Sway with Self-Occlusion using Top-View Camera

Abstract

We propose a method for identifying people using body sway measured from head regions under a condition that a top-view camera observes bodies disturbed by self-occlusion. In order to correctly represent identities of body sway, it is necessary that the appearances of people are accurately acquired from the camera. However, the deficits of the appearances are contained when occuring self-occlusion. The deficits cause a problem that the identification performance of an existing method is wrongly degraded because the method uses whole-body regions to represent identities. To overcome the problem of self-occlusion, we represent identities of body sway by using head regions acquired from a top-view camera. Our method computes the silhouette images of the head regions by applying a segmentation technique. To represent identities of body sway, we temporally measure the movements of the head regions by spatially dividing them into local regions. The experimental results show that the identification performance improved from 17.3 % using the whole-body regions of the existing method to 57.9 % using the head regions of our method.