ISCA Archive ASVspoof 2024
ISCA Archive ASVspoof 2024

Intema system description for the ASVspoof5 Challenge: power weighted score fusion

Ali Aliyev, Alexander Kondratev

The paper describes Intema submissions to the ASVspoof 5 challenge for Track 1 deepfake detection and Track 2 SASV system (both open). Independent spoofing countersure (CM) and automatic speaker verification (ASV) systems based on Self-supervised learning (SSL) and ResNet-like models were developed. The training process is clearly described for CM and ASV tasks. An aggregation method for this systems is proposed. The final submission for the Track 1 (open) achieved minDCF 0.0936, combined system for Track 2 (open) achieved min a-DCF 0.1203.