Automatic Voice Identification after Speech Resynthesis using PPG

TL;DR

This paper introduces a PPG-based speech resynthesis system that maintains audio quality and demonstrates that speaker identity cannot be recovered after resynthesis, enhancing privacy and speaker anonymization.

Contribution

The paper presents a novel PPG-based speech resynthesis approach and evaluates its effectiveness in preventing speaker identification post-resynthesis.

Findings

Resynthesis produces high-quality audio.

Speaker verification fails to identify source speaker after resynthesis.

PPG effectively disentangles speaker identity from phonetic content.

Abstract

Speech resynthesis is a generic task for which we want to synthesize audio with another audio as input, which finds applications for media monitors and journalists.Among different tasks addressed by speech resynthesis, voice conversion preserves the linguistic information while modifying the identity of the speaker, and speech edition preserves the identity of the speaker but some words are modified.In both cases, we need to disentangle speaker and phonetic contents in intermediate representations.Phonetic PosteriorGrams (PPG) are a frame-level probabilistic representation of phonemes, and are usually considered speaker-independent.This paper presents a PPG-based speech resynthesis system.A perceptive evaluation assesses that it produces correct audio quality.Then, we demonstrate that an automatic speaker verification model is not able to recover the source speaker after re-synthesis with PPG, even when the model is trained on synthetic data.