Additive Margin in Contrastive Self-Supervised Frameworks to Learn Discriminative Speaker Representations

TL;DR

This paper introduces an Additive Margin to contrastive SSL methods for speaker verification, improving embedding discriminability and reducing errors, leading to state-of-the-art performance.

Contribution

It proposes the NT-Xent-AM loss with Additive Margin and demonstrates their effectiveness in enhancing SSL speaker representations.

Findings

Additive Margin improves embedding compactness.

Reduces false negatives and positives in speaker verification.

Achieves 7.85% EER on VoxCeleb1-O, outperforming previous methods.

Abstract

Self-Supervised Learning (SSL) frameworks became the standard for learning robust class representations by benefiting from large unlabeled datasets. For Speaker Verification (SV), most SSL systems rely on contrastive-based loss functions. We explore different ways to improve the performance of these techniques by revisiting the NT-Xent contrastive loss. Our main contribution is the definition of the NT-Xent-AM loss and the study of the importance of Additive Margin (AM) in SimCLR and MoCo SSL methods to further separate positive from negative pairs. Despite class collisions, we show that AM enhances the compactness of same-speaker embeddings and reduces the number of false negatives and false positives on SV. Additionally, we demonstrate the effectiveness of the symmetric contrastive loss, which provides more supervision for the SSL task. Implementing these two modifications to SimCLR improves performance and results in 7.85% EER on VoxCeleb1-O, outperforming other equivalent methods.