My final talk focuses on the integration of biometric systems with spoofing countermeasures. The first two editions of ASVspoof focused upon the assessment of countermeasures in isolation from automatic speaker verification systems; the talk will explain the reasons behind this strategy. With spoofing countermeasures always operating alongside a biometric recogniser, however, a more meaningful approach to assessment should reflect the intertwined operation and impact of each separate component on the single, combined system as a whole. The talk discusses different solutions to the integration of automatic speaker verification and spoofing countermeasures and shows how the detection error cost function used for the assessment of automatic speaker verification systems can be extended to assess their performance when they are combined with spoofing countermeasures.
Automatic speaker verification and spoofing countermeasures: Integration and metrics
S4P Summer School on Speech Signal Processing, Speaker Recognition and Diarization, DAIICT, 6-10 July 2019, Gandhinagar, India
© EURECOM. Personal use of this material is permitted. The definitive version of this paper was published in S4P Summer School on Speech Signal Processing, Speaker Recognition and Diarization, DAIICT, 6-10 July 2019, Gandhinagar, India and is available at :
PERMALINK : https://www.eurecom.fr/publication/6163