eduzhai > Applied Sciences > Engineering >

Bunched LPCNet Vocoder for Low-cost Neural Text-To-Speech Systems

  • king
  • (0) Download
  • 20210506
  • Save

... pages left unread,continue reading

Document pages: 5 pages

Abstract: LPCNet is an efficient vocoder that combines linear prediction and deepneural network modules to keep the computational complexity low. In this work,we present two techniques to further reduce it s complexity, aiming for alow-cost LPCNet vocoder-based neural Text-to-Speech (TTS) System. Thesetechniques are: 1) Sample-bunching, which allows LPCNet to generate more thanone audio sample per inference; and 2) Bit-bunching, which reduces thecomputations in the final layer of LPCNet. With the proposed bunchingtechniques, LPCNet, in conjunction with a Deep Convolutional TTS (DCTTS)acoustic model, shows a 2.19x improvement over the baseline run-time whenrunning on a mobile device, with a less than 0.1 decrease in TTS mean opinionscore (MOS).

Please select stars to rate!

         

0 comments Sign in to leave a comment.

    Data loading, please wait...
×