eduzhai > Applied Sciences > Engineering >

VAW-GAN for Singing Voice Conversion with Non-parallel Training Data

  • king
  • (0) Download
  • 20210506
  • Save

... pages left unread,continue reading

Document pages: 6 pages

Abstract: Singing voice conversion aims to convert singer s voice from source to targetwithout changing singing content. Parallel training data is typically requiredfor the training of singing voice conversion system, that is however notpractical in real-life applications. Recent encoder-decoder structures, such asvariational autoencoding Wasserstein generative adversarial network (VAW-GAN),provide an effective way to learn a mapping through non-parallel training data.In this paper, we propose a singing voice conversion framework that is based onVAW-GAN. We train an encoder to disentangle singer identity and singing prosody(F0 contour) from phonetic content. By conditioning on singer identity and F0,the decoder generates output spectral features with unseen target singeridentity, and improves the F0 rendering. Experimental results show that theproposed framework achieves better performance than the baseline frameworks.

Please select stars to rate!

         

0 comments Sign in to leave a comment.

    Data loading, please wait...
×