Datasets and codes

Data

• ASVspoof 5 Evaluation Plan
• ASVspoof 2019 LA Listening Test Data for Partial Rank Similarity MOS Prediction
• Voice Conversion Challenge 2020 database
• Corpus of Age-related Voice DisguiseCorpus of Age-related Voice Disguise (AVOID)
• Automatic Speaker Verification Spoofing and Countermeasures Challenge (ASVspoof). See [IEEE-J-STSP overview paper of the ASVspoof challenge]
  • ASVspoof 2021 challenge, webpage, workshop, data[LA][PA][DF], metadata[LA][PA][DF], baseline systems
  • ASVSpoof2019 challenge data and webpage
  • ASVSpoof2019 “Real PA” set
  • ASVspoof2017 challenge data (audio replay attack detection)  [NOTE! this is patched v2.0 of the corpus, described here and recommended to be used instead of the original one] See also ICASSP 2017 paper about data collection  and Interspeech 2017 challenge overview paper
  • ASVspoof2015 challenge data (voice conversion and text-to-speech attack detection task).
• The Voice Conversion Challenge 2018: database and results (VCC18). See also the challenge overview paper and another paper containing supplementary speech artifact analysis (both will be presented at Odyssey 2018)
• I-vectors (~420 MB) used in IEEE-T-IFS paper (hosted at IDIAP).
• I4U consortium filelists for NIST SRE12 development purposes (from Rahim Saeidi’s pages) used in [Interspeech 2013 paper]

Program codes

• t-EER: Parameter-Free Tandem Evaluation of Countermeasures and Biometric Comparators, and Jupyter Notebook
• Household ASV baseline system, described in [Speaker Odyssey workshop paper]
• ASVTorch toolkit, described in ASVtorch toolkit: Speaker verification with deep neural networks
• Code for training an agent to play Minecraft by learning from human demonstrations.
• Program code for reading the image of a Windows/Linux game window and emulating keyboard/mouse controls.
• Toribash learning environment for training agents in a hand-to-hand combat setup.
• GPU accelerated implementation of i-vector extractor (training / extraction) using PyTorch.
• Semi-supervised speech activity detector, described in [Computer Speech & Language paper]
• Audio replay attack detection baseline code (Matlab), for ASVspoof 2017 challenge
• Local variability features (Matlab). See [the related paper in Digital Signal Processing]
• PLDA for anti-spoofing (Python). See also [IEEE-T-IFS paper]
• Fast probabilistic linear discriminant analysis (PLDA) implementation (Matlab and Python). See the related S+SSPR paper [PDF]
• Utterance-by-utterance adaptive speech activity detector (SAD) presented in ICASSP 2013 [PDF].
• Multiple window (multitaper) spectrum estimators (Matlab). See also the related publications in IEEE T-ASLP, Speech Communication, IEEE SPL,  Interspeech 2010 and ASRU 2011.
• Regularized all-pole methods as an appendix of Odyssey 2012 paper. See also the related publication in IEEE SPL.
• Temporally weighted linear predictors (from Jouni Pohjalainen’s page)