Full video processing for mobile audio-visual identity verification

Alexander Usoltsev (1, 2, 3) , Dijana Petrovska-Delacrétaz (4, 1, 2, 3) , Houssemeddine Khemiri (1, 2, 3)

1 TSP - EPH - Département Electronique et Physique
2 SAMOVAR - Services répartis, Architectures, MOdélisation, Validation, Administration des Réseaux
3 CNRS - Centre National de la Recherche Scientifique
4 ARMEDIA-SAMOVAR - ARMEDIA

This paper describes a bi-modal biometric verification system based on voice and face modalities, which takes advantage of the full video processing instead of using still-images. The bi-modal system is evaluated on the MOBIO corpus and results show a relative improvement of performance by nearly 10% when the whole video is used. The fusion between face and speaker verification systems, using linear logistic regression weights, gives a relative improvement of performance that varies between 30% and 60% comparing to the best uni-modal system. Proof-of-concept iPad application is developed based on the proposed bi-modal system

Mots clés: Biometrics; Face; Speech; Full video processing; Score fusion

Domaines: Traitement du signal et de l'image [eess.SP]

Alexander Usoltsev, Dijana Petrovska-Delacrétaz, Houssemeddine Khemiri. Full video processing for mobile audio-visual identity verification. ICPRAM 2016 : 5th International Conference on Pattern Recognition Applications and Methods, Feb 2016, Rome, Italy. pp.552 - 557, ⟨10.5220/0005667305520557⟩. ⟨hal-01451049⟩