eduzhai > Applied Sciences > Engineering >

Nonparallel Voice Conversion with Augmented Classifier Star Generative Adversarial Networks

  • king
  • (0) Download
  • 20210507
  • Save

... pages left unread,continue reading

Document pages: 14 pages

Abstract: We previously proposed a method that allows for nonparallel voice conversion(VC) by using a variant of generative adversarial networks (GANs) calledStarGAN. The main features of our method, called StarGAN-VC, are as follows:First, it requires no parallel utterances, transcriptions, or time alignmentprocedures for speech generator training. Second, it can simultaneously learnmappings across multiple domains using a single generator network and thusfully exploit available training data collected from multiple domains tocapture latent features that are common to all the domains. Third, it cangenerate converted speech signals quickly enough to allow real-timeimplementations and requires only several minutes of training examples togenerate reasonably realistic-sounding speech. In this paper, we describe threeformulations of StarGAN, including a newly introduced novel StarGAN variantcalled "Augmented classifier StarGAN (A-StarGAN) ", and compare them in anonparallel VC task. We also compare them with several baseline methods.

Please select stars to rate!


0 comments Sign in to leave a comment.

    Data loading, please wait...