Skip to main content

Table 6 Results of systems using hybrid multi-task learning with different configurations on VoxCeleb

From: Introducing phonetic information to speaker embedding for speaker verification

Ā 

EER(%)

minDCF08

minDCF10

x-vector

2.68

0.0144

0.4645

x-vector-mt (1-layer sharing)

2.58

0.0132

0.4027

x-vector-mt (2-layer sharing)

2.73

0.0145

0.3977

x-vector-mt (3-layer sharing)

2.83

0.0151

0.4700

x-vector-mt (4-layer sharing)

2.92

0.0151

0.5001