Emo-StarGAN: A Semi-Supervised Any-to-Many Non-Parallel Emotion-Preserving Voice Conversion

TL;DR

This paper introduces Emo-StarGAN, a semi-supervised voice conversion model that effectively preserves emotion during anonymization, using emotion-aware losses and classifier supervision to improve over existing methods.

Contribution

It presents a novel semi-supervised any-to-many voice conversion approach that enhances emotion preservation while maintaining anonymization on non-parallel data.

Findings

Significant improvement in emotion preservation demonstrated

Effective across diverse datasets and emotions

Maintains intelligibility and anonymization

Abstract

Speech anonymisation prevents misuse of spoken data by removing any personal identifier while preserving at least linguistic content. However, emotion preservation is crucial for natural human-computer interaction. The well-known voice conversion technique StarGANv2-VC achieves anonymisation but fails to preserve emotion. This work presents an any-to-many semi-supervised StarGANv2-VC variant trained on partially emotion-labelled non-parallel data. We propose emotion-aware losses computed on the emotion embeddings and acoustic features correlated to emotion. Additionally, we use an emotion classifier to provide direct emotion supervision. Objective and subjective evaluations show that the proposed approach significantly improves emotion preservation over the vanilla StarGANv2-VC. This considerable improvement is seen over diverse datasets, emotions, target speakers, and inter-group conversions without compromising intelligibility and anonymisation.