Logo des Repositoriums

Auflistung nach Autor:in "Uhl,Andreas"

1 - 10 von 29
  • Textdokument
    Benchmarking Fingerprint Minutiae Extractors
    (BIOSIG 2017, 2017) Chugh,Tarang; Arora,Sunpreet S.; Jain, Anil K.; Paulter Jr.,Nicholas G.
    The performance of a fingerprint recognition system hinges on the errors introduced in each of its modules: image acquisition, preprocessing, feature extraction, and matching. One of the most critical and fundamental steps in fingerprint recognition is robust and accurate minutiae extraction. Hence we conduct a repeatable and controlled evaluation of one open-source and three commercial-off-the-shelf (COTS) minutiae extractors in terms of their performance in minutiae detection and localization. We also evaluate their robustness against controlled levels of image degradations introduced in the fingerprint images. Experiments were conducted on (i) a total of 3;458 fingerprint images from five public-domain databases, and (ii) 40;000 synthetically generated fingerprint images. The contributions of this study include: (i) a benchmark for minutiae extractors and minutiae interoperability, and (ii) robustness of minutiae extractors against image degradations.
  • Textdokument
    Biometric Systems under Morphing Attacks: Assessment of Morphing Techniques and Vulnerability Reporting
    (BIOSIG 2017, 2017) Scherhag,Ulrich; Nautsch,Andreas; Rathgeb,Christian; Gomez-Barrero,Marta; Veldhuis,Raymond N.J.; Spreeuwers,Luuk; Schils,Maikel; Maltoni,Davide; Grother,Patrick; Marcel,Sébastien; Breithaupt,Ralph; Ramachandra,Raghavendra; Busch,Christoph
    With the widespread deployment of biometric recognition systems, the interest in attacking these systems is increasing. One of the easiest ways to circumvent a biometric recognition system are so-called presentation attacks, in which artefacts are presented to the sensor to either impersonate another subject or avoid being recognised. In the recent past, the vulnerabilities of biometric systems to so-called morphing attacks have been unveiled. In such attacks, biometric samples of multiple subjects are merged in the signal or feature domain, in order to allow a successful verification of all contributing subjects against the morphed identity. Being a recent area of research, there is to date no standardised manner to evaluate the vulnerability of biometric systems to these attacks. Hence, it is not yet possible to establish a common benchmark between different morph detection algorithms. In this paper, we tackle this issue proposing new metrics for vulnerability reporting, which build upon our joint experience in researching this challenging attack scenario. In addition, recommendations on the assessment of morphing techniques and morphing detection metrics are given.
  • Textdokument
    BIOSIG 2017
    (BIOSIG 2017, 2017)
  • Textdokument
    De-duplication using automated face recognition: a mathematical model and all babies are equally cute
    (BIOSIG 2017, 2017) Spreeuwers,Luuk
    De-duplication is defined as the technique to eliminate or link duplicate copies of repeating data. We consider a specific de-duplication application where a subject applies for a new passport and we want to check if he possesses a passport already under another name. To determine this, a facial photograph of the subject is compared to all photographs of the national database of passports.We investigate if state of the art facial recognition is up to this task and find that for a large database about 2 out of 3 duplicates can be found while few or no false duplicates are reported. This means that de-duplication using automated face recognition is feasible in practice.We also present a mathematical model to predict the performance of de-duplication and find that the probability that k false duplicates are returned can be described well by a Poisson distribution using a varying, subject specific false match rate. We present experimental results using a large database of actual passport photographs consisting of 224 000 images of about 100 000 subjects and find that the results are predicted well by our model.
  • Textdokument
    Deep Quality-informed Score Normalization for Privacy-friendly Speaker Recognition in unconstrained Environments
    (BIOSIG 2017, 2017) Nautsch,Andreas; Steen,Søren Trads; Busch,Christoph
    In scenarios that are ambitious to protect sensitive data in compliance with privacy regulations, conventional score normalization utilizing large proportions of speaker cohort data is not feasible for existing technology, since the entire cohort data would need to be stored on each mobile device. Hence, in this work we motivate score normalization utilizing deep neural networks. Considering unconstrained environments, a quality-informed scheme is proposed, normalizing scores depending on sample quality estimates in terms of completeness and signal degradation by noise. Utilizing the conventional PLDA score, comparison i-vectors, and corresponding quality vectors, we aim at mimicking cohort based score normalization optimizing the Cmin llr discrimination criterion. Examining the I4U data sets for the 2012 NIST SRE, an 8.7% relative gain is yielded in a pooled 55-condition scenario with a corresponding condition-averaged relative gain of 6.2% in terms of Cmin llr . Robustness analyses towards sensitivity regarding unseen conditions are conducted, i.e. when conditions comprising lower quality samples are not available during training.
  • Textdokument
    Domain Adaptation for CNN Based Iris Segmentation
    (BIOSIG 2017, 2017) Jalilian,Ehsaneddin; Uhl,Andreas; Kwitt,Roland
    Convolutional Neural Networks (CNNs) have shown great success in solving key artificial vision challenges such as image segmentation. Training these networks, however, normally requires plenty of labeled data, while data labeling is an expensive and time-consuming task, due to the significant human effort involved. In this paper we propose two pixel-level domain adaptation methods, introducing a training model for CNN based iris segmentation. Based on our experiments, the proposed methods can effectively transfer the domains of source databases to those of the targets, producing new adapted databases. The adapted databases then are used to train CNNs for segmentation of iris texture in the target databases, eliminating the need for the target labeled data. We also indicate that training a specific CNN for a new iris segmentation task, maintaining optimal segmentation scores, is possible using a very low number of training samples.
  • Textdokument
    Evaluation of CNN architectures for gait recognition based on optical flow maps
    (BIOSIG 2017, 2017) Castro,Francisco M.; Marín-Jiménez,Manuel J.; Guil,Nicolás; López-Tapia,Santiago; de la Blanca,Nicolás Pérez
    This work targets people identification in video based on the way they walk (i.e.gait) by using deep learning architectures. We explore the use of convolutional neural networks (CNN) for learning high-level descriptors from low-level motion features (i.e.optical flow components). The low number of training samples for each subject and the use of a test set containing subjects different from the training ones makes the search of a good CNN architecture a challenging task.We carry out a thorough experimental evaluation deploying and analyzing four distinct CNN models with different depth but similar complexity. We show that even the simplest CNN models greatly improve the results using shallow classifiers. All our experiments have been carried out on the challenging TUMGAID dataset, which contains people in different covariate scenarios (i.e.clothing, shoes, bags).
  • Textdokument
    Evaluation of Motion-based Touch-typing Biometrics in Online Financial Environments
    (BIOSIG 2017, 2017) Buriro,Attaullah; Gupta,Sandeep; Crispo,Bruno
    This paper presents a bimodal scheme, the mechanism which contemplates the way a user enters an 8-digit PIN/password and the phone-movements while doing so, for user authentication in mobile banking/financial applications (apps). The scheme authenticates the user based on the timing differences of the entered strokes. Additionally, it enhances the security by introducing a transparent layer utilizing the phone-movements made by the user. The scheme is assumed to be highly secure as mimicking the invisible touch-timings and the phone-movements could be extremely onerous. Our analysis is based on 2850 samples collected from 95 users through a 3-day unsupervised field experiment and using 3 multi-class classifiers. Random Forest (RF) classifier out-performed other two classifiers and provided a True Acceptance Rate (TAR) of 96%.
  • Textdokument
    Exploring Texture Transfer Learning via Convolutional Neural Networks for Iris Super Resolution
    (BIOSIG 2017, 2017) Ribeiro,Eduardo; Uhl,Andreas
    Increasingly, iris recognition towards more relaxed conditions has issued a new superresolution field direction. In this work we evaluate the use of deep learning and transfer learning for single image super resolution applied to iris recognition. For this purpose, we explore if the nature of the images as well as if the pattern from the iris can influence the CNN transfer learning and, consequently, the results in the recognition process. The good results obtained by the texture transfer learning using a deep architecture suggest that features learned by Convolutional Neural Networks used for image super-resolution can be highly relevant to increase iris recognition rate.
  • Textdokument
    Fingerprint Damage Localizer and Detector of Skin Diseases from Fingerprint Images
    (BIOSIG 2017, 2017) Barotova,Stepanka; Drahansky,Martin
    This article describes a novel approach for detection and classification of skin diseases in fingerprints using three methods - Block Orientation Field, Histogram Analysis and Flood Fill. The combination of these methods brings a surprising results and using a rule descriptor for selected skin diseases, we are able to classify the disease into a group or concrete name.