The 18th edition of the International Conference of the Biometrics Special Interest Group (BIOSIG) took place at Fraunhofer IGD located in Darmstadt, Germany from September 18 to 20 and attracted 84 registered participants.
The program was composed of scientific research contributions and three keynote talks. The first keynote talk was given by Lena Klasen from the Swedish forensic laboratory. After briefly summarizing the history of forensics, she gave an insight of crime prediction in Sweden (shootings, explosions, etc.). Further, she described an identification approach based on video surveillance data with person description and face recognition. Finally, she presented a planned multi-modal system to be used for forensics investigations employing different sensors for different scenarios.
In the first research contribution a online survey with more than 700 participants was conducted in order estimate the public perceptions and preferences towards ATMs with biometric authentication in Austria. The study did not reveal strong majority preferences. The usage of cards together with PINs is still preferred but could be augmented with biometrics.
Subsequently, a session on the topic of face morphing gave insights on the vulnerability of face recognition towards attacks based on morphed face. It was shown that a separate weighting of warping and blending factors can be used to optimize to success rate of an attacker. Further, it was demonstrated that deep neural style transfer can be applied for improving the quality of morphed faces.
The next session featured three research contribution related to eye-based biometric recognition. Firstly, a presentation attack detection method based on adversarial learning was presented. Particular emphasis was put on the generalization to unseen attacks. Secondly, a new deep-learning based iris segmentation scheme was introduced. Finally, it was shown that a strong correlation between left and right retinal images of single person exists. The results were based on image data acquired with a special device of the Brno university, which simultaneously captures the iris and retina of a data subject.
The last research session of the first day focused on fingerprint recognition. On the one hand, a new deep-learning-based approach for fingerprint pre-alignment using Siamese networks was presented. On the other hand a novel 2400 dpi finger acquisition device was introduced which can be used to capture neonate fingerprints. Results from a field study in Kenia were shown.
Then, further conference contributions were presented in the poster session. The topics of the 9 posters ranged from homomorphic encryption for face identification to gender recognition in visual surveillance environments. The poster session got great reception from the attendees who afterwards joined the traditional barbeque leading to a lot of opportunities for networking and technical discussions.
The last day started with a keynote talk by Takashi Shinzaki from the Fujitsu research laboratory in Japan. He provided interesting details about the Fujitsu palm vein technology and discussed many relevant aspects including efficient one-to-many comparison and template protection.
Following the keynote speech, a research work on finger vein recognition was presented tackling the problem of longitudinal finger rotation during finger vein acquisition. Afterwards a multi-spectral presentation attack detection approach was presented which makes use of laser speckle contrast imaging on fingerprints with a multi-algorithm feature extraction.
In a final research session works on the application of deep-learning for behavioural biometrics (keystrokes and gait recognition) as well as clothes recognition in urban environment have been presented.
Eventually, Walter Scheirer from the University of Notre Dame discussed the limits and potentials of deep learning for facial analysis in his well-attended keynote talk. He showed experiments on the effects of different image processing, such as blur, contrast and rotation, on the recognition accuracy of different deep face recognition. Obtained results were compared with human recognition capabilities. It was demonstrated that visual psychophysics is a viable methodology for making face recognition algorithms more explainable. In the second part of his talk, the potential of deep-learning for describable visual facial attribute modelling was explored.
Attendees of the conference voted for the best paper and best poster presented during the conference which was honored with a prize and a certificate. Participants decided for the work of Yoshinori Koda (NEC Corporation) “Development of 2,400ppi Fingerprint Sensor for Capturing Neonate Fingerprint within 24 Hours after Birth”. The winning poster was presented by Pawel Drozdowski with the topic “On the Application of Homomorphic Encryption to Face Identification”.
The BIOSIG conference was preceded by several events: the meeting of the TeleTrusT Biometric Working Group, the EAB Research Project Conference, the EAB Biometrics Research and Industry Award and the general assembly of the European Association of Biometrics.
The 2019 BIOSIG conference was jointly organized by the Competence Center for Applied Security Technology (CAST) and the special interest group BIOSIG of the Gesellschaft für Informatik e.V. (GI).
The conference was technically co-sponsored by IEEE Biometric Council and the papers will be added to IEEE Xplore.
Next year the BIOSIG conference will take place between
September 16 to 18, 2020 in Darmstadt, Germany.
Look on our page for updates in the near future: www.biosig.org