Advances in Biometrics: International Conference, ICB 2006, by Kieron Messer, Josef Kittler, James Short, G. Heusch, Fabien PDF

By Kieron Messer, Josef Kittler, James Short, G. Heusch, Fabien Cardinaux (auth.), David Zhang, Anil K. Jain (eds.)

This booklet constitutes the refereed lawsuits of the foreign convention on Biometrics, ICB 2006, held in Hong Kong, China in January 2006.

The 104 revised complete papers provided have been rigorously reviewed and chosen from 192 submissions. Biometric standards coated by way of the papers are assigned to stand, fingerprint, iris, speech and signature, biometric fusion and function assessment, gait, keystrokes, and others. furthermore the result of the Face Authentication festival (FAC 2006) also are introduced during this volume.

Show description

Read or Download Advances in Biometrics: International Conference, ICB 2006, Hong Kong, China, January 5-7, 2006. Proceedings PDF

Best computers books

S. Raghavan, G. Anandalingam's Telecommunications Planning: Innovations in Pricing, Network PDF

The examine awarded in Telecommunications making plans: ideas in Pricing, community layout and administration specializes in the newest methodological advancements in 3 key components – pricing of telecommunications prone, community layout, and source allocation. those 3 components are so much appropriate to present telecommunications making plans.

Additional info for Advances in Biometrics: International Conference, ICB 2006, Hong Kong, China, January 5-7, 2006. Proceedings

Sample text

For a review of the GMM-UBM models, we refer the readers to the work in [8] that has been successfully applied in speaker verification. During testing, as shown in Figure 2(b), given a test image, we test if the image belongs to the class of images with good quality or poor quality. This is achieved using the Maximum Likelihood decision rule. We applied this approach to the Cohn-Kanade database [9]. Our experiments show that the accuracy of the system is 75% in discriminating between the images with good quality and the images with poor quality.

This raises a new research question: is it possible to create a training set that will be representative for the whole world? If not, then the face recognition system used in border control might be population dependent: people who do not belong to the target population, from which the training data set is selected, will probably generate higher number of false acceptances compared to people who belong to the target population. The eye coordinates of the passport photos in the data set III were generated automatically, which means that the 16 % of the eye coordinates were not correct.

Additional information about the experiment can be found in MSc thesis of M. Kosmerlj [8]. Face Recognition Issues in a Border Control Environment 3 37 The Effect of Additional Human Inspection Based on our discoveries of look-alikes that might be able to pass a computerized face recognition environment, a natural next step would be to investigate whether an additional human guard would detect these false acceptances by the computerized system. In the previous experiment the computerized face recognition system compared normalized images without hair while in a real-world situation the people passing a control post will have hair.

Download PDF sample

Rated 4.35 of 5 – based on 13 votes