HLTCOE Submission to the VoicePrivacy Attacker Challenge - Improving ASV Performance Against Anonymized Speech

This 15-minute conference talk from the Center for Language & Speech Processing (CLSP) at Johns Hopkins University details their submission to the 2024 VoicePrivacy Attacker Challenge. Learn about three main categories of methods developed to improve Automatic Speaker Verification (ASV) performance against anonymized speech: classifier improvements, alternative distance metrics for ASV score computation, and kNN-VC normalization. Discover how the implementation of these techniques, either individually or in combination, resulted in significant reductions in Equal Error Rate (EER) when tested against all anonymization systems submitted to the VoicePrivacy Challenge.