This baseline build on well-established iVector/PLDA speaker verification framework for SRE 2010 female tasks. This work is based on Kaldi SRE 2008 example. The result is reported as the clean iVector/PLDA baseline in [1] with average EER 3.50% over nine core conditions.
To run this experiment plese first check data, hardware and software requirements and then read this guide to repeat experment.
For information about kaldi IO please read this document.
If you feel this baseline helpful kindly cite Kaldi as well as this paper:
[1] Steven Du, Xiong Xiao, Eng Siong Chng, "DNN FEATURE COMPENSATION FOR NOISE ROBUST SPEAKER VERIFICATION", ChinaSIP 2015