eduzhai > Applied Sciences > Engineering >

Textual Echo Cancellation

  • king
  • (0) Download
  • 20210506
  • Save

... pages left unread,continue reading

Document pages: 8 pages

Abstract: In this paper, we propose Textual Echo Cancellation (TEC) - a framework forcancelling the text-to-speech (TTS) playback echo from overlapped speechrecordings. Such a system can largely improve speech recognition performanceand user experience for intelligent devices such as smart speakers, as the usercan talk to the device while the device is still playing the TTS signalresponding to the previous query. We implement this system by using a novelsequence-to-sequence model with multi-source attention that takes both themicrophone mixture signal and the source text of the TTS playback as inputs,and predicts the enhanced audio. Experiments show that the textual informationof the TTS playback is critical to the enhancement performance. Besides, thetext sequence is much smaller in size compared with the raw acoustic signal ofthe TTS playback, and can be immediately transmitted to the device and the ASRserver even before the playback is synthesized. Therefore, our proposedapproach effectively reduces Internet communication and latency compared withalternative approaches such as acoustic echo cancellation (AEC).

Please select stars to rate!


0 comments Sign in to leave a comment.

    Data loading, please wait...