eduzhai > Applied Sciences > Engineering >

Accurate Detection of Wake Word Start and End Using a CNN

  • king
  • (0) Download
  • 20210506
  • Save

... pages left unread,continue reading

Document pages: 5 pages

Abstract: Small footprint embedded devices require keyword spotters (KWS) with smallmodel size and detection latency for enabling voice assistants. Such a keywordis often referred to as textit{wake word} as it is used to wake up voiceassistant enabled devices. Together with wake word detection, accurateestimation of wake word endpoints (start and end) is an important task of KWS.In this paper, we propose two new methods for detecting the endpoints of wakewords in neural KWS that use single-stage word-level neural networks. Ourresults show that the new techniques give superior accuracy for detecting wakewords endpoints of up to 50 msec standard error versus human annotations, onpar with the conventional Acoustic Model plus HMM forced alignment. To ourknowledge, this is the first study of wake word endpoints detection methods forsingle-stage neural KWS.

Please select stars to rate!

         

0 comments Sign in to leave a comment.

    Data loading, please wait...
×