Skip to content

Latest commit

 

History

History
20 lines (14 loc) · 963 Bytes

README.md

File metadata and controls

20 lines (14 loc) · 963 Bytes
Raw

x-vector-anonymization-using-autoencoder-and-adversial-training-for-preserving-speech-privacy

this repository contains pytorch implementation of article "X-vector anonymization using autoencoders and adversarial training for preserving speech privacy"

jornal = Computer Speech & Language (CSL)

year = 2022

authors = Juan M. Perero-Codosero, Fernando M. Espinoza-Cuadros, Luis A. Hernández-Gómez b

link to original paper

Notes :

  • The authors of this article used x-vector but I used d-vector. They are all the same and are used for speaker embedding.
  • I just implement the vector anonymization part
  • I train on a much smaller dataset from Kaggle instead of VoxCeleb which is a huge dataset
  • You only need an Kaggle.com account and request for an API key for downloading the dataset

Language and framework used :

  • Python
  • Pytorch