A compact ResNet trained with GE2E or AAMSoftmax identifies the source TTS system of synthetic speech as accurately as a 315M-parameter Wav2Vec2/AASIST model on MLAADv5.
The ResNet model exhibited mixed perfor- mance when trained with different classification loss functions
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.SD 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Synthetic Speech Source Tracing using Metric Learning
A compact ResNet trained with GE2E or AAMSoftmax identifies the source TTS system of synthetic speech as accurately as a 315M-parameter Wav2Vec2/AASIST model on MLAADv5.