Median f0 and HNR source-output consistency detect degraded voice clones at 80-85% accuracy on WaveRNN and HiFi-GAN samples.
Among the three features,f 0 and HNR were consistently the most informa- tive, while VTL was weaker across both vocoders
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
citation-role summary
background 1
citation-polarity summary
fields
eess.AS 1years
2026 1verdicts
UNVERDICTED 1roles
background 1polarities
background 1representative citing papers
citing papers explorer
-
Low-Cost Detection of Degraded Voice Clones via Source-Output Acoustic Consistency
Median f0 and HNR source-output consistency detect degraded voice clones at 80-85% accuracy on WaveRNN and HiFi-GAN samples.