1. experiments
  2. pkorshunov
  3. pkorshunov/isv-speaker-verification/1
  4. isv-speaker-verification-licit

ISV modeling for speaker verification on AVspoof dataset (full genuine data)

Outputs for block analysis

eer 0.0479927
far_dev 0.0479381
far_test 0.052294
frr_dev 0.0480473
frr_test 0.0314359
hter 0.0418649
threshold 0.485038
number_of_negatives_dev 54925
number_of_negatives_test 70620
number_of_positives_dev 4225
number_of_positives_test 4708
roc_dev
roc_test
To edit a toolchain, please use a modern browser (Mozilla Firefox 3.0+, Google Chrome 1+, Apple Safari 3+, Opera 9.5+, Microsoft Internet Explorer 9+)

This is an automatic speaker verification (ASV) system based on ISV modeling for a typical licit scenario. In this scenario, there are no spoofing attacks present, only genuine zero-impostors. Full AVspoof database is used in this experiment. GMM with 256 Gaussians was used to train a UBM.

Experiments which are referred by reports or by an associated attestation cannot be modified. You may revert the situation by deleting unpublished reports or the attestation in case they have not been published. Note you may still fork this experiment and re-run a modified or unmodified version of it.

This experiment is not yet attested.
Updated Name Actions
July 15, 2016 pkorshunov/ISV-based-ASV-system-licit-scenario (Performance of ISV-based ASV system on data with no attacks)
1
Terms of Service | Contact Information | BEAT platform version 2.2.1b0 | © Idiap Research Institute - 2013-2026