Enhancing Zero-Shot Many to Many Voice Conversion with Self-Attention VAE

TL;DR

This paper enhances VAE-based zero-shot many-to-many voice conversion by integrating self-attention and regularization techniques, leading to improved speaker classification accuracy and voice quality on unseen speakers.

Contribution

It introduces a novel combination of self-attention placement and relaxed group-wise splitting regularization to improve VAE performance in voice conversion.

Findings

28.3% increase in speaker classification accuracy on unseen speakers

Slight improvement in voice quality as per MOSNet scores

Effective incorporation of non-local information in VAE decoder

Abstract

Variational auto-encoder (VAE) is an effective neural network architecture to disentangle a speech utterance into speaker identity and linguistic content latent embeddings, then generate an utterance for a target speaker from that of a source speaker. This is possible by concatenating the identity embedding of the target speaker and the content embedding of the source speaker uttering a desired sentence. In this work, we propose to improve VAE models with self-attention and structural regularization (RGSM). Specifically, we found a suitable location of VAE's decoder to add a self-attention layer for incorporating non-local information in generating a converted utterance and hiding the source speaker's identity. We applied relaxed group-wise splitting method (RGSM) to regularize network weights and remarkably enhance generalization performance. In experiments of zero-shot many-to-many voice conversion task on VCTK data set, with the self-attention layer and relaxed group-wise splitting method, our model achieves a gain of speaker classification accuracy on unseen speakers by 28.3\% while slightly improved conversion voice quality in terms of MOSNet scores. Our encouraging findings point to future research on integrating more variety of attention structures in VAE framework while controlling model size and overfitting for advancing zero-shot many-to-many voice conversions.