HIGHLIGHTS
- What: The authors provide further insights influence anonymized between human perception and automatic similarity scoring. Through this work, the authors aim to inform how evaluation techniques and datasets may be designed in the future. The authors study how to characterize speakers based on their performance in a set of voice anonymization systems from the 2020 VoicePrivacy Challenge . The authors focus this work on one type of attack model in voice privacy called the lazy-informed attack model.
- Who: Jennifer Williams and collaborators from the University of Southampton, UK have published the article: Anonymizing Speaker Voices . . .
If you want to have access to all the content you need to log in!
Thanks :)
If you don't have an account, you can create one here.