da/sec scientific talk on Biometrics

Topic: Feature-type fusion for mobile iris recognition

by Esbern Andersen-Hoppe
FBI D19/2.03, Mai 12, 2016 (Thursday), 12.00 noon

Keywords — Biometrics, Mobile iris recognition, Fusion, Feature extraction, Visual wavelength


The use of mobile devices has been increasing drasticly over the last few years and they are being used more and more to store and access data. Some of that data might need to be protected in a way where a simple PIN-number will be insufficient. Seeing as iris recognition in the visual spectrum has received more attention over the years it opens up the possibility of using the camera of a mobile device for iris recognition to access data. However, the camera in a mobile device is usually not able to capture high quality images which means a feature extractor might not be able to extract enough information to securely distinguish between an imposter and a genuine access attempt. Since some feature extractors collect different types of information a fusion of different types of feature extractors might complement each other to increase the possibility to distinguish between an imposter and a genuine access attempt.

The presentation will show the results of fusing several different types of feature extractors and how the recognition performance increases. Experiments have been made on two publicly available mobile databases (UBIRIS v2 and MobBIO) where ground truth information have been available such that incorrect iris segmentation would not be an issue.