Table 5 Results obtained by sharing different numbers of layers in hybrid multi-task learning
From: Introducing phonetic information to speaker embedding for speaker verification
EER(%) | minDCF08 | minDCF10 | |
|---|---|---|---|
x-vector | 1.96 | 0.0109 | 0.3900 |
x-vector-mt (1-layer sharing) | 1.67 | 0.0091 | 0.3465 |
x-vector-mt (2-layer sharing) | 1.61 | 0.0082 | 0.3009 |
x-vector-mt (3-layer sharing) | 1.52 | 0.0073 | 0.2752 |
x-vector-mt (4-layer sharing) | 1.50 | 0.0086 | 0.3383 |