Skip to main content

Table 15 Summary of results on the CMN2 subsets of NIST SRE 2018. The actDCF18 is also reported on the evaluation set

From: Introducing phonetic information to speaker embedding for speaker verification

Ā 

Dev

Eval

Ā 

EER(%)

minDCF18

EER(%)

minDCF18

actDCF18

x-vector

8.86

0.587

10.94

0.630

0.631

x-vector-pv (c=0)

8.10

0.556

9.92

0.597

0.604

x-vector-pv (c=0.2)

7.82

0.521

9.62

0.557

0.562

x-vector-mt (1-layer sharing)

8.40

0.554

10.07

0.589

0.593

sc-vector (1-layer sharing)

8.35

0.543

10.06

0.585

0.587

c-vector (c=0.2 + 1-layer sharing)

7.95

0.511

9.57

0.579

0.584