Title:Egocentric Video Biometrics

Abstract:Egocentric cameras are being worn by an increasing number of users, among them many security forces worldwide. GoPro cameras already penetrated the mass market, and Google Glass may follow soon. As head-worn cameras do not capture the face and body of the wearer, it may seem that the anonymity of the wearer can be preserved even when the video is publicly distributed.
We show that motion features in egocentric video provide biometric information, and the identity of the user can be reliably determined from a few seconds of video captured when the user is walking. The proposed method achieves more than 90% identification accuracy in cases where the random success rate is only 3%.
Applications may include theft prevention of wearable cameras by locking the camera when not worn by its lawful owner. This work can also provide the first steps towards searching on video sharing services (e.g. YouTube) for egocentric videos shot by a specific person. An important message in this paper is that people should be aware that sharing egocentric video will compromise their anonymity, even when their face is not visible.