Janusz Konrad | ECE Department, Boston University | July 21 th, 2014, 14:00 CET | Lakeside Labs, Room L4.1.114
Abstract: How many times per day do you need to prove your identity? Perhaps 5, maybe 10 or even 20. Try to access your smartphone and you need a 4-digit PIN at minimum. Need to check email at work and you have to provide username and password. Want to enter a lab and you need to punch in a code, swipe a magnetic card or wave an RFID tag. While cards and tags can be easily lost, codes and passwords are difficult to remember once we have too many of them, and we all do! One answer to this dilemma is to use biometrics such as fingerprints, irises or facial characteristics, but should they be compromised there is no way to replace them. In this talk, I will describe our recent work on using soft biometrics jointly with gestures to authenticate a user. We implicitly use rough proportions of human body shape from skeleton data captured by Kinect camera. Should these soft biometrics become compromised, a user can easily change his/her gesture and prevent the loss of identity. I will describe details of our approach and some experimental results in which we have attained a 1-2% equal error rate on a database of 40 users. We believe this level of performance shows promise for swinging your way to security in a not-too-distant future.
Bio: Janusz Konrad earned the M.Eng. degree from the Technical University of Szczecin, Poland, and the Ph.D. degree from McGill University, Montreal, Canada. From 1989 to 2000 he was with INRS-Telecommunications, Montreal, and since 2000 with Boston University. He has been on Editorial Boards of various IEEE and EURASIP journals related to signal, image and video processing. He was the General Chair of AVSS-2013, a Technical Program Co-Chair of ICIP-2000 and AVSS-2010, and Tutorials Co-Chair of ICASSP-2004. He is a co-recipient of the 2001 Signal Processing Magazine Award, the 2004-05 EURASIP Image Communications Best Paper Award, the AVSS-2010 Best Paper Award and a co-winner of the Semantic Description of Human Activities Contest at ICPR-2010. His research interests include image and video processing, stereoscopic and 3D displays, human-computer interfaces and visual sensor networks.
