Language-Independent Speaker Anonymization Approach using Self-Supervised Pre-Trained Models

TL;DR

This paper introduces a simplified, language-independent speaker anonymization method using self-supervised pre-trained models, effectively protecting speaker privacy across different languages without relying on language-dependent components.

Contribution

The paper proposes a novel SSL-based speaker anonymization approach that eliminates the need for language-specific models, enabling easy adaptation to multiple languages.

Findings

Effective anonymization in English and Mandarin datasets

Simpler model architecture compared to existing methods

Maintains speech intelligibility and privacy protection

Abstract

Speaker anonymization aims to protect the privacy of speakers while preserving spoken linguistic information from speech. Current mainstream neural network speaker anonymization systems are complicated, containing an F0 extractor, speaker encoder, automatic speech recognition acoustic model (ASR AM), speech synthesis acoustic model and speech waveform generation model. Moreover, as an ASR AM is language-dependent, trained on English data, it is hard to adapt it into another language. In this paper, we propose a simpler self-supervised learning (SSL)-based method for language-independent speaker anonymization without any explicit language-dependent model, which can be easily used for other languages. Extensive experiments were conducted on the VoicePrivacy Challenge 2020 datasets in English and AISHELL-3 datasets in Mandarin to demonstrate the effectiveness of our proposed SSL-based language-independent speaker anonymization method.