eduzhai > Applied Sciences > Engineering >

End-to-end spoofing detection with raw waveform CLDNNs

  • king
  • (0) Download
  • 20210505
  • Save

... pages left unread,continue reading

Document pages: 5 pages

Abstract: Albeit recent progress in speaker verification generates powerful models,malicious attacks in the form of spoofed speech, are generally not coped with.Recent results in ASVSpoof2015 and BTAS2016 challenges indicate thatspoof-aware features are a possible solution to this problem. Most successfulmethods in both challenges focus on spoof-aware features, rather than focusingon a powerful classifier. In this paper we present a novel raw waveform baseddeep model for spoofing detection, which jointly acts as a feature extractorand classifier, thus allowing it to directly classify speech signals. Thisapproach can be considered as an end-to-end classifier, which removes the needfor any pre- or post-processing on the data, making training and evaluation astreamlined process, consuming less time than other neural-network basedapproaches. The experiments on the BTAS2016 dataset show that the systemperformance is significantly improved by the proposed raw waveformconvolutional long short term neural network (CLDNN), from the previous bestpublished 1.26 half total error rate (HTER) to the current 0.82 HTER.Moreover it shows that the proposed system also performs well under the unknown(RE-PH2-PH3,RE-LPPH2-PH3) conditions.

Please select stars to rate!

         

0 comments Sign in to leave a comment.

    Data loading, please wait...
×