[Subject Prev][Subject Next][Thread Prev][Thread Next][Date Index][Thread Index]

[hts-users:04434] ASVspoof 2017


=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*

 ASVspoof 2017 CHALLENGE:
 Audio replay detection for automatic speaker verification anti-spoofing

 http://www.spoofingchallenge.org/

=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*

 Are you good at machine learning for audio signals? Are you good at
 discriminating 'fake' signals from authentic ones? Are you looking for new
 audio processing challenges? Do you work in the domain of speaker recognition?
 ASVspoof 2017 challenge might be for you!

CHALLENGE TASK:

 Given a short clip of speech audio, determine whether it contains
 a GENUINE human voice (live recording), or a REPLAY recording (fake).

 You will be provided a development set containing genuine/replay labeled audio
 examples, along with further metadata such as speech content and the devices
 used in the replay recordings. Your task is to develop a system that assigns a
 single 'liveness' or 'genuineness' score value to new audio samples, and to
 execute that system on a set of test files for which the ground truth is not
 provided. We provide a Matlab-based reference baseline method to kick-off
 quickly towards developing your new ideas!

 For more details, refer to the evaluation plan in the website:
 http://www.spoofingchallenge.org/

BACKGROUND:

 The goal of the challenge series is to enhance security of automatic speaker
 verification (ASV) systems from being intentially circumvented using fake
 recordings, also known as 'spoofing attacks' or 'representation attacks' in
 the context of biometrics. ASVspoof 2017 is a second edition of a challenge
 kicked off in 2015, and the new perspective in ASVspoof 2017 are the replay
 attacks, especially 'unseen' attacks - for instance, containing replay
 environments, devices and speakers that might be very different from those in
 the development data.

 Despite 'ASV' being in the challenge title, you do NOT require knowledge
 of automatic speaker verification: the task is a 'standalone' replay audio
 detection task that can be addressed as a generic acoustic pattern classification
 problem. We welcome as many new ideas to the problem as possible!

SCHEDULE:

 Development data published:   December 23th, 2016
 Evaluation data published:    February 10, 2017
 Evaluation set scores due:    February 24, 2017
 Results available:            March 3, 2017
 Interspeech paper deadline:   March 14, 2017
 Metadata/keys published:      May 2017
 Interspeech special session:  August 2017

REGISTRATION:

 Send a free-worded e-mail to asvspoof2017@xxxxxxxxx
 to register and obtain the dev data.

ORGANIZERS:

 Tomi Kinnunen, University of Eastern Finland, FINLAND
 Nicholas Evans, Eurecom, FRANCE
 Junichi Yamagishi, University of Edinburgh, UK
 Kong Aik Lee, Institute for Infocomm Research, SINGAPORE
 Md Sahidullah, University of Eastern Finland, FINLAND
 Massimiliano Todisco, Eurecom, FRANCE
 Hector Delgado, Eurecom, FRANCE

CONTACT:

 asvspoof2017@xxxxxxxxx



The University of Edinburgh is a charitable body, registered in
Scotland, with registration number SC005336.