Clustering speaker voiceprints and packing training batches with within-cluster negative pairs improves supervised contrastive speaker verification by up to 18% relative EER on VoxCeleb.
For a fair comparison, in all experiments, we use the exact same model, data and training parameters
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
eess.AS 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Clustering-based hard negative sampling for supervised contrastive speaker verification
Clustering speaker voiceprints and packing training batches with within-cluster negative pairs improves supervised contrastive speaker verification by up to 18% relative EER on VoxCeleb.