MOBIO D3.2: Report on the description and evaluation of baseline algorithms for unimodal authentication

speech recognition, speaker authentication, unimodal system, face verification, silence detection, speaker verification

This report is on the description and evaluation of baseline algorithms for unimodal authentication.

This deliverable specifies the technological overview of current biometric baseline systems, both state-of-the-art unimodal face and speaker authentication approaches. The deliverable will elaborate the main scientific contributions, details of implementation, evaluation and results on baseline face authentication and speaker authentication through combining techniques of face localisation, facial feature localisation & face verification and silence detection & speaker verification respectively. These baseline algorithms are evaluated independently and as well tested together as unimodal authentication systems using the well-known BANCA database. The evaluation of these baseline algorithms/systems will: a) provide a basis for developing new methodologies to enhance the performance of current unimodal authentication systems, b) help in innovative development of fusion schemes for integrating unimodal systems to form robust bi-modal authentication systems and c) finally allow comparisons with these joint authentication methods.