Font Size: a A A

Multi-modal biometrics fusion for human recognition in video

Posted on:2008-05-07Degree:Ph.DType:Dissertation
University:University of California, RiversideCandidate:Zhou, XiaoliFull Text:PDF
GTID:1448390005969000Subject:Engineering
Abstract/Summary:
Biometrics deal with uniquely recognizing humans based upon one or more intrinsic physical or behavioral traits. Even though the performance of a single biometrics system is constrained by the intrinsic factors of a trait, this inherent limitation of a single biometric can be alleviated by fusing the information presented by multiple sources. It has been found to be difficult to recognize a person from arbitrary views when one is walking at a distance. In this dissertation, we propose an innovative video based fusion system, which aims at recognizing non-cooperating individuals at a distance in a single camera scenario by integrating information of the side view of face and gait.; Firstly, we propose a new face representation, Enhanced Side Face Image (ESFI), a higher resolution image compared with the image directly obtained from a single video frame, for a side face. It overcomes the problem of the limited resolution of face at a distance in video by integrating information from multiple video frames. The experimental results show that the idea of constructing ESFI from multiple frames is promising for human recognition in video and more discriminating face features are extracted from ESFI compared to those from the original side face images (OSFI).; Secondly, we explore several schemes to fuse the information of the side view of face and gait at two different levels: the match score level and the feature level. Performance comparisons between different fusion methods are presented. The performance is also shown in Cumulative Match Characteristic (CMC) curves. The experimental results show that the integration of information from side face and gait is effective for individual recognition in video.; Finally, we develop a non-reference objective measure to evaluate the quality of the super-resolved face images constructed under different conditions. Face recognition experiments are also conducted based on the super-resolved images. Experimental results demonstrate that the proposed quality measure is effective in the quality assessment of super-resolved images and that the image resolution enhancement is necessary for the recognition task in the low resolution image/video scenario.
Keywords/Search Tags:Video, Recognition, Face, Fusion, Images, Resolution
Related items