Challenging margin-based speaker embedding extractors by using the variational information bottleneck

Challenging margin-based speaker embedding extractors by using the variational information bottleneck

Paper in proceedings (conference paper)

Speaker embedding extractors are typically trained using a classification loss over the training speakers. During the last few years, the standard softmax/cross-entropy loss has been replaced by the margin-based losses, yielding significant im- provements in speaker recognition accuracy. Motivated by the fact that the margin merely reduces the logit of the target speaker during training, we consider a probabilistic framework that has a similar effect. The variational information bottle- neck provides a principled mechanism for making deterministic nodes stochastic, resulting in an implicit reduction of the pos- terior of the target speaker. We experiment with a wide range of speaker recognition benchmarks and scoring methods and re- port competitive results to those obtained with the state-of-the- art Additive Angular Margin loss.

Speaker embedding extractors are typically trained using a classification loss over the training speakers. During the last few years, the standard softmax/cross-entropy loss has been replaced by the margin-based losses, yielding significant im- provements in speaker recognition accuracy. Motivated by the fact that the margin merely reduces the logit of the target speaker during training, we consider a probabilistic framework that has a similar effect. The variational information bottle- neck provides a principled mechanism for making deterministic nodes stochastic, resulting in an implicit reduction of the pos- terior of the target speaker. We experiment with a wide range of speaker recognition benchmarks and scoring methods and re- port competitive results to those obtained with the state-of-the- art Additive Angular Margin loss.

speaker recognition, variational information bottleneck

speaker recognition, variational information bottleneck

STAFYLAKIS, T.; SILNOVA, A.; ROHDIN, J.; PLCHOT, O.; BURGET, L.

2025

01.09.2024

International Speech Communication Association

Kos

Proceedings of Interspeech 2024

1990-9772

Proceedings of Interspeech

2024

9

French Republic

3220

3224

5

https://www.isca-archive.org/interspeech_2024/stafylakis24_interspeech.pdf

stafylakis_interspeech_2024