eduzhai > Applied Sciences > Engineering >

Speaker-Conditional Chain Model for Speech Separation and Extraction

  • Save

... pages left unread,continue reading

Document pages: 7 pages

Abstract: Speech separation has been extensively explored to tackle the cocktail partyproblem. However, these studies are still far from having enough generalizationcapabilities for real scenarios. In this work, we raise a common strategy namedSpeaker-Conditional Chain Model to process complex speech recordings. In theproposed method, our model first infers the identities of variable numbers ofspeakers from the observation based on a sequence-to-sequence model. Then, ittakes the information from the inferred speakers as conditions to extract theirspeech sources. With the predicted speaker information from whole observation,our model is helpful to solve the problem of conventional speech separation andspeaker extraction for multi-round long recordings. The experiments fromstandard fully-overlapped speech separation benchmarks show comparable resultswith prior studies, while our proposed model gets better adaptability formulti-round long recordings.

Please select stars to rate!

         

0 comments Sign in to leave a comment.

    Data loading, please wait...
×