eduzhai > Applied Sciences > Engineering >

Few-Shot Keyword Spotting With Prototypical Networks

  • king
  • (0) Download
  • 20210505
  • Save

... pages left unread,continue reading

Document pages: 6 pages

Abstract: Recognizing a particular command or a keyword, keyword spotting has beenwidely used in many voice interfaces such as Amazon s Alexa and Google Home. Inorder to recognize a set of keywords, most of the recent deep learning basedapproaches use a neural network trained with a large number of samples toidentify certain pre-defined keywords. This restricts the system fromrecognizing new, user-defined keywords. Therefore, we first formulate thisproblem as a few-shot keyword spotting and approach it using metric learning.To enable this research, we also synthesize and publish a Few-shot GoogleSpeech Commands dataset. We then propose a solution to the few-shot keywordspotting problem using temporal and dilated convolutions on prototypicalnetworks. Our comparative experimental results demonstrate keyword spotting ofnew keywords using just a small number of samples.

Please select stars to rate!

         

0 comments Sign in to leave a comment.

    Data loading, please wait...
×