MotivationMethod of modeling SNR informationi-vector6dB6 dBSNRWSubspacecln15 dBcleanW15dBI-vectorSpace
Motivation 6 dB • Method of modeling SNR information clean 15 dB SNR Subspace w6dB w cln w15dB I-vector Space i-vector
Distribution of SNR in SRE12SNRdistributionof test utterances inCC4formale350EachSNRregionishandledbya300specificsetof SNRfactors25020015010050o01020305040SNR/dB
Distribution of SNR in SRE12 Each SNR region is handled by a specific set of SNR factors
Contents1.Background2.MotivationofWork3.SNR-invariant PLDA modeling for Robust SpeakerVerification4.ExperimentsonSRE125. Conclusions
Contents 1. Background 2. Motivation of Work 3. SNR-invariant PLDA modeling for Robust Speaker Verification 4. Experiments on SRE12 5. Conclusions