README.md

Understanding the Tradeoffs in Client-side Privacy for Downstream Speech Tasks

Here is a link to our generated samples.

Notes

• Link to GAN voice conversion model.

• Link to VAE voice conversion model.

• We used REAPER and Rubber Band to build our signal processing approach. Details are described in our paper.

• Run src/main.py to train the adversary classifier.

• Link to ASR model.

Paper

Understanding the Tradeoffs in Client-side Privacy for Downstream Speech Tasks

@inproceedings{wu2021privacy,
      title={Understanding the Tradeoffs in Client-side Privacy for Downstream Speech Tasks},
      author={Peter Wu and Paul Pu Liang and Jiatong Shi and Ruslan Salakhutdinov and Shinji Watanabe and Louis-Philippe Morency},
      booktitle={APSIPA ASC},
      year={2021}
}