eduzhai > Applied Sciences > Engineering >

Cross attentive pooling for speaker verification

  • king
  • (0) Download
  • 20210506
  • Save

... pages left unread,continue reading

Document pages: 7 pages

Abstract: The goal of this paper is text-independent speaker verification whereutterances come from in the wild videos and may contain irrelevant signal.While speaker verification is naturally a pair-wise problem, existing methodsto produce the speaker embeddings are instance-wise. In this paper, we proposeCross Attentive Pooling (CAP) that utilizes the context information across thereference-query pair to generate utterance-level embeddings that contain themost discriminative information for the pair-wise matching problem. Experimentsare performed on the VoxCeleb dataset in which our method outperformscomparable pooling strategies.

Please select stars to rate!


0 comments Sign in to leave a comment.

    Data loading, please wait...