AASIST3: KAN-Enhanced AASIST Speech Deepfake Detection using SSL Features and Additional Regularization for the ASVspoof 2024 Challenge
AASIST3 améliore la détection de deepfakes vocaux en intégrant des réseaux de Kolmogorov-Arnold (KAN) au framework AASIST. Le modèle atteint minDCF=0.5357 (closed) et 0.1414 (open) sur ASVspoof 2024, doublant les performances précédentes. Code disponible sur HuggingFace.