SVSNet+: Enhancing Speaker Voice Similarity Assessment Models with Representations from Speech Foundation Models

TL;DR

This paper introduces SVSNet+, a model that leverages pre-trained speech foundation model representations, like WavLM, to significantly enhance speaker voice similarity assessment accuracy across multiple datasets.

Contribution

SVSNet+ is the first to systematically incorporate pre-trained speech foundation model representations into speaker similarity assessment, demonstrating improved performance and generalization.

Findings

SVSNet+ with WavLM outperforms baseline models on Voice Conversion Challenge datasets.

Learning a weighted-sum of WavLM features improves performance more than fine-tuning.

SVSNet+ maintains strong performance with different speech foundation models.

Abstract

Representations from pre-trained speech foundation models (SFMs) have shown impressive performance in many downstream tasks. However, the potential benefits of incorporating pre-trained SFM representations into speaker voice similarity assessment have not been thoroughly investigated. In this paper, we propose SVSNet+, a model that integrates pre-trained SFM representations to improve performance in assessing speaker voice similarity. Experimental results on the Voice Conversion Challenge 2018 and 2020 datasets show that SVSNet+ incorporating WavLM representations shows significant improvements compared to baseline models. In addition, while fine-tuning WavLM with a small dataset of the downstream task does not improve performance, using the same dataset to learn a weighted-sum representation of WavLM can substantially improve performance. Furthermore, when WavLM is replaced by other SFMs, SVSNet+ still outperforms the baseline models and exhibits strong generalization ability.