Skip to content

StevenLOL/Research_speech_speaker_verification_nist_sre2010

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

34 Commits
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Research_speech_speaker_verification_nist_sre2010 [working in progress]

This baseline build on well-established iVector/PLDA speaker verification framework for SRE 2010 female tasks. It is based on Kaldi SRE2008 example,forked under version r3473.

#There are two setups for SRE2010 with different using of SRE2008 data ##Setup 1 SRE2008 is not used for Tmatrix,PLDA,LDA training, reported as the clean iVector/PLDA base line in [1] with average EER 3.50%. Results reported by other researchers on SRE2010 in papers[?] ###UBM training Switchboard II Phase 2 and 3, Switchboard Cellular Part 1 and 2, and NIST SRE 2004, 2005 and 2006 ###I Vector Extractor training Switchboard II Phase 2 and 3, Switchboard Cellular Part 1 and 2, and NIST SRE 2004, 2005 and 2006 ###Development Dataset SRE2008 ###Test Dataset SRE2010

##Setup 2 SRE2008 is used for Tmatrix,PLDA,LDA training (WIP),results reported by other researchers in this setup[??] ###UBM training Switchboard II Phase 2 and 3, Switchboard Cellular Part 1 and 2, and NIST SRE 2004, 2005 and 2006,SRE2008 ###I Vector Extractor training Switchboard II Phase 2 and 3, Switchboard Cellular Part 1 and 2, and NIST SRE 2004, 2005 and 2006,SRE2008 ###Development Dataset XX ###Test Dataset SRE2010

#Prerequisites ##data having the Data listed above ##NIST DATA ##NIST Keys #Kaldi #Install Kaldi #Kaldi structure

If you use this baseline kindly cite paper:

[1] Steven Du, Xiong Xiao, Eng Siong Chng, "DNN FEATURE COMPENSATION FOR NOISE ROBUST SPEAKER VERIFICATION", in proceedings of ChinaSIP 2015

About

Research_speech_speaker_verification_nist_sre2010

Topics

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published