SEED refines noisy speaker embeddings toward clean ones using a diffusion-style training objective, improving EER by up to 19.6% on a simulated mismatch benchmark with no speaker labels.
Title resolution pending
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
-
SEED: Speaker Embedding Enhancement Diffusion Model
SEED refines noisy speaker embeddings toward clean ones using a diffusion-style training objective, improving EER by up to 19.6% on a simulated mismatch benchmark with no speaker labels.