Delivering Speaking Style in Low-resource Voice Conversion with Multi-factor Constraints

TL;DR

This paper introduces MFC-StyleVC, a novel low-resource voice conversion model that effectively preserves speaking style and target speaker timbre using multi-factor constraints and a simulation mode, outperforming existing methods.

Contribution

The paper proposes a new VC model with a clustering-based timbre constraint, perceptual regularization, and a simulation mode to improve low-resource voice conversion performance.

Findings

Outperforms existing low-resource VC methods in expressive speech tasks.

Effectively maintains speaking style, content, and quality with limited data.

Demonstrates robustness through extensive experiments.

Abstract

Conveying the linguistic content and maintaining the source speech's speaking style, such as intonation and emotion, is essential in voice conversion (VC). However, in a low-resource situation, where only limited utterances from the target speaker are accessible, existing VC methods are hard to meet this requirement and capture the target speaker's timber. In this work, a novel VC model, referred to as MFC-StyleVC, is proposed for the low-resource VC task. Specifically, speaker timbre constraint generated by clustering method is newly proposed to guide target speaker timbre learning in different stages. Meanwhile, to prevent over-fitting to the target speaker's limited data, perceptual regularization constraints explicitly maintain model performance on specific aspects, including speaking style, linguistic content, and speech quality. Besides, a simulation mode is introduced to simulate the inference process to alleviate the mismatch between training and inference. Extensive experiments performed on highly expressive speech demonstrate the superiority of the proposed method in low-resource VC.