Eta-WavLM: Efficient Speaker Identity Removal in Self-Supervised Speech Representations Using a Simple Linear Equation

TL;DR

This paper introduces Eta-WavLM, a simple linear method to effectively disentangle speaker identity from speech representations in self-supervised learning, improving performance in content-focused tasks like voice conversion.

Contribution

The paper presents a novel linear decomposition technique for disentangling speaker information from SSL speech representations, enhancing speaker independence without complex models.

Findings

Achieves effective speaker disentanglement in SSL representations.

Improves voice conversion performance over existing methods.

Maintains content integrity while removing speaker information.

Abstract

Self-supervised learning (SSL) has reduced the reliance on expensive labeling in speech technologies by learning meaningful representations from unannotated data. Since most SSL-based downstream tasks prioritize content information in speech, ideal representations should disentangle content from unwanted variations like speaker characteristics in the SSL representations. However, removing speaker information often degrades other speech components, and existing methods either fail to fully disentangle speaker identity or require resource-intensive models. In this paper, we propose a novel disentanglement method that linearly decomposes SSL representations into speaker-specific and speaker-independent components, effectively generating speaker disentangled representations. Comprehensive experiments show that our approach achieves speaker independence and as such, when applied to content-driven tasks such as voice conversion, our representations yield significant improvements over state-of-the-art methods.