Towards Supervised Performance on Speaker Verification with Self-Supervised Learning by Leveraging Large-Scale ASR Models

TL;DR

This paper demonstrates that fine-tuning large-scale SSL models like WavLM with pseudo-labels can significantly improve speaker verification performance, approaching supervised system accuracy.

Contribution

It introduces a novel SSL-based framework using pseudo-labels and iterative refinement to enhance speaker representations in WavLM, achieving state-of-the-art results.

Findings

Achieved 0.99% EER on VoxCeleb1-O, setting new state-of-the-art in self-supervised SV.

Close to supervised baseline performance with 0.94% EER.

Proved the effectiveness of pseudo-label refinement in SSL speaker verification.

Abstract

Recent advancements in Self-Supervised Learning (SSL) have shown promising results in Speaker Verification (SV). However, narrowing the performance gap with supervised systems remains an ongoing challenge. Several studies have observed that speech representations from large-scale ASR models contain valuable speaker information. This work explores the limitations of fine-tuning these models for SV using an SSL contrastive objective in an end-to-end approach. Then, we propose a framework to learn speaker representations in an SSL context by fine-tuning a pre-trained WavLM with a supervised loss using pseudo-labels. Initial pseudo-labels are derived from an SSL DINO-based model and are iteratively refined by clustering the model embeddings. Our method achieves 0.99% EER on VoxCeleb1-O, establishing the new state-of-the-art on self-supervised SV. As this performance is close to our supervised baseline of 0.94% EER, this contribution is a step towards supervised performance on SV with SSL.