- TextdokumentFingerprint Template Ageing vs. Template Changes Revisited(BIOSIG 2017, 2017) Kirchgasser,Simon; Uhl,Andreas; Brömme,Arslan; Busch,Christoph; Dantcheva,Antitza; Rathgeb,Christian; Uhl,AndreasThis study investigates the impact of “ghost” fingerprint and minutiae information in 4 year time-span separated fingerprint datasets. A high amount of ghost fingerprints within the data, eventually a source for differences in acquisition conditions, might be responsible for recently reported template ageing effects. According to that, various experiments have been performed to get rid of this problematic image content and to compare the corresponding matching results to the performance figures using the non altered imprints. The analysis with respect to detected increased error rates exhibits very similar effects for all considered methods no matter if ghost fingerprint information is removed or not. Thus, ghost fingerprints are not responsible for the observed effects.
- TextdokumentRecognizing infants and toddlers over an on-production fingerprint database(BIOSIG 2017, 2017) Camacho,Vanina; Garella,Guillermo; Franzoni,Francesco; Di Martino,Luis; Carbajal,Guillermo; Preciozzi,Javier; Fernández,Alicia; Brömme,Arslan; Busch,Christoph; Dantcheva,Antitza; Rathgeb,Christian; Uhl,AndreasIt is widely known that biometric systems based on adults fingerprints have reached an outstanding performance when compared against other biometric traits. This explains their extensive use by governmental agencies in charge of citizen identification. Nevertheless, the performance is highly degraded when fingerprints of newborns or toddlers are used. In this work, we analyze the performance of existing solutions (both at sensor and matching level) using 45000 infants fingerprints taken from an on-production civilian database. We also propose a solution by zooming the input fingerprints with an interpolation factor based on ridges distances. The developed solution shows improvements in both fingerprint quality (NFIQ 2.0) as well as recognition performance.
- TextdokumentxTARP: Improving the Tented Arch Reference Point Detection Algorithm(BIOSIG 2017, 2017) Merkle,Johannes; Tams,Benjamin; Dieckmann,Benjamin; Korte,Ulrike; Brömme,Arslan; Busch,Christoph; Dantcheva,Antitza; Rathgeb,Christian; Uhl,AndreasIn 2013, Tams et al. proposed a method to determine directed reference points in fingerprints based on a mathematical model of typical orientation fields of tented arch type fingerprints. Although this Tented Arch Reference Point (TARP) method has been used successfully for prealignment in biometric cryptosystems, its accuracy does not yet ensure satisfactory error rates for single finger systems. In this paper, we improve the TARP algorithm by deploying an improved orientation field computation and by integrating an additional mathematical model for arch type fingerprints. The resulting Extended Tented Arch Reference Point (xTARP) method combines the arch model with the tented arch model and achieves a significantly better accuracy than the original TARP algorithm. When deploying the xTARP method in the Fuzzy Vault construction of Butt et al., the false non-match rate (FNMR) at a security level of 20 bits is reduced from 7:4% to 1:7%.
- TextdokumentImprovement of Iris Recognition based on Iris-Code Bit-Error Pattern Analysis(BIOSIG 2017, 2017) Rathgeb,Christian; Busch,Christoph; Brömme,Arslan; Busch,Christoph; Dantcheva,Antitza; Rathgeb,Christian; Uhl,AndreasIn this paper an advanced iris-biometric comparator is presented. In the proposed scheme an analysis of bit-error patterns produced by Hamming distance-based iris-code comparisons is performed. The lengths of sequences of horizontal consecutive mis-matching bits are measured and a frequency distribution is estimated. The difference of the extracted frequency distribution to that of an average genuine one obtained from a training set is used as a second comparison score. This score is then used together with the fractional Hamming distance in order to improve the recognition accuracy of an iris recognition system. In experimental evaluations relative improvements of approximately 45% and 10% in terms of false non-match rate at a false match rate of 0.01% are achieved on the CASIAv4-Interval and the BioSecure iris databases, respectively.
- TextdokumentSIC-Gen: A Synthetic Iris-Code Generator(BIOSIG 2017, 2017) Drozdowski,Pawel; Rathgeb,Christian; Busch,Christoph; Brömme,Arslan; Busch,Christoph; Dantcheva,Antitza; Rathgeb,Christian; Uhl,AndreasNowadays large-scale identity management systems enrol more than one billion data subjects. In order to limit transaction times, biometric indexing is a suitable method to reduce the search space in biometric identifications. Effective testing of such biometric identification systems and biometric indexing approaches requires large datasets of biometric data. Currently, the size of the publicly available iris datasets is insufficient, especially for system scalability assessments. Synthetic data generation offers a potential solution to this issue; however, it is challenging to generate data hat is both statistically sound and visually realistic - for the iris, the currently available approaches prove unsatisfactory. In this paper, we present a method for generation of synthetic binary iris-based templates, i.e. Iris-Codes, which are the de facto standard used throughout major biometric deployments around the world. We validate the statistical properties of the synthetic templates and show that they closely resemble ones produced from real ocular images. With the proposed approach, large databases of synthetic Iris-Codes with flexibly adjustable properties can be generated.
- TextdokumentDomain Adaptation for CNN Based Iris Segmentation(BIOSIG 2017, 2017) Jalilian,Ehsaneddin; Uhl,Andreas; Kwitt,Roland; Brömme,Arslan; Busch,Christoph; Dantcheva,Antitza; Rathgeb,Christian; Uhl,AndreasConvolutional 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.
- TextdokumentBIOSIG 2017(BIOSIG 2017, 2017) Brömme,Arslan; Busch,Christoph; Dantcheva,Antitza; Rathgeb,Christian; Uhl,Andreas
- TextdokumentHow Random is a Classifier given its Area under Curve?(BIOSIG 2017, 2017) Zeinstra,Chris; Veldhuis,Raymond; Spreeuwers,Luuk; Brömme,Arslan; Busch,Christoph; Dantcheva,Antitza; Rathgeb,Christian; Uhl,AndreasWhen the performance of a classifier is empirically evaluated, the Area Under Curve (AUC) is commonly used as a one dimensional performance measure. In general, the focus is on good performance (AUC towards 1). In this paper, we study the other side of the performance spectrum (AUC towards 0.50) as we are interested to which extend a classifier is random given its AUC. We present the exact probability distribution of the AUC of a truely random classifier, given a finite number of distinct genuine and imposter scores. It quantifies the “randomness” of the measured AUC. The distribution involves the restricted partition function, a well studied function in number theory. Although other work exists that considers confidence bounds on the AUC, the novelty is that we do not assume any underlying parametric or non-parametric model or specify an error rate. Also, in cases in which a limited number of scores is available, for example in forensic case work, the exact distribution can deviate from these models. For completeness, we also present an approximation using a normal distribution and confidence bounds on the AUC.
- TextdokumentEvaluation 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; Brömme,Arslan; Busch,Christoph; Dantcheva,Antitza; Rathgeb,Christian; Uhl,AndreasThis 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).
- TextdokumentFusing Biometric Scores using Subjective Logic for Gait Recognition on Smartphone(BIOSIG 2017, 2017) Wasnik,Pankaj; Schäfer,Kirstina; Raja,Kiran; Ramachandra,Raghavendra; Busch,Christoph; Brömme,Arslan; Busch,Christoph; Dantcheva,Antitza; Rathgeb,Christian; Uhl,AndreasThe performance of a biometric system gets affected by various types of errors such as systematic errors, random errors, etc. These kinds of errors usually occur due to the natural variations in the biometric traits of subjects, different testing, and comparison methodologies. Neither of these errors can be easily quantifiable by mathematical formulas. This behavior introduces an uncertainty in the biometric verification or identification scores. The combination of comparison scores from different comparators or combination of multiple biometric modalities could be a better approach for improving the overall recognition performance of a biometric system. In this paper, we propose a method for combining such scores from multiple comparators using Subjective Logic (SL), as it takes uncertainty into account while performing to biometric fusion. This paper proposes a framework for a smartphone based gait recognition system with application of SL for biometric data fusion.