Automatic video-based person authentication using the RBF network

Abstract

As more and more forensic information becomes available on video we address in this paper the Automatic Video-Based Biometric Person Authentication (AVBPA). Possible tasks and application scenarios under consideration involve detection and tracking of humans and human (ID) verification. Authentication corresponds to ID verification and involves actual (face) recognition for the subject(s) detected in the video sequence. The architecture for AVBPA takes advantage of the active vision paradigm and it involves difference methods or optical flow analysis to detect the moving subject, projection analysis and decision trees (DT) for face location, and connectionist network — Radial Basis Function (RBF) for authentication. Subject detection and face location correspond to video break and key frame detection, respectively, while recognition itself corresponds to authentication. The active vision paradigm is most appropriate for video processing where one has to cope with huge amounts of image data and where further sensing and processing of additional frames is feasible. As a result of such an approach video processing becomes feasible in terms of decreased computational resources (‘time’) spent and increased confidence in the (authentication) decisions reached despite sometime poor quality imagery. Experimental results on three FERET video sequences prove the feasibility of our approach.