Anonymizing Speech with Generative Adversarial Networks to Preserve Speaker Privacy
This work addresses privacy protection for individuals in speech data, though it appears incremental as it builds on existing anonymization methods.
The paper tackled the problem of speaker anonymization in speech data by generating artificial speaker embeddings using a generative adversarial network with Wasserstein distance, resulting in improved privacy and utility over previous approaches as confirmed by objective metrics and human evaluation.
In order to protect the privacy of speech data, speaker anonymization aims for hiding the identity of a speaker by changing the voice in speech recordings. This typically comes with a privacy-utility trade-off between protection of individuals and usability of the data for downstream applications. One of the challenges in this context is to create non-existent voices that sound as natural as possible. In this work, we propose to tackle this issue by generating speaker embeddings using a generative adversarial network with Wasserstein distance as cost function. By incorporating these artificial embeddings into a speech-to-text-to-speech pipeline, we outperform previous approaches in terms of privacy and utility. According to standard objective metrics and human evaluation, our approach generates intelligible and content-preserving yet privacy-protecting versions of the original recordings.