eduzhai > Applied Sciences > Engineering >

Source-Aware Neural Speech Coding for Noisy Speech Compression

  • king
  • (0) Download
  • 20210507
  • Save

... pages left unread,continue reading

Document pages: 5 pages

Abstract: This paper introduces a novel neural network-based speech coding system thatcan process noisy speech effectively. The proposed source-aware neural audiocoding (SANAC) system harmonizes a deep autoencoder-based source separationmodel and a neural coding system so that it can explicitly perform sourceseparation and coding in the latent space. An added benefit of this system isthat the codec can allocate a different amount of bits to the underlyingsources so that the more important source sounds better in the decoded signal.We target a new use case where the user on the receiver side cares about thequality of the non-speech components in speech communication, while the speechsource still carries the most crucial information. Both objective andsubjective evaluation tests show that SANAC can recover the original noisyspeech better than the baseline neural audio coding system, which is with nosource-aware coding mechanism, and two conventional codecs.

Please select stars to rate!


0 comments Sign in to leave a comment.

    Data loading, please wait...