From: Multi-task deep cross-attention networks for far-field speaker verification and keyword spotting
 | Distance | Seen | Unseen | ||||
---|---|---|---|---|---|---|---|
 | EER (%) | minDCF | Acc (%) | EER (%) | minDCF | Acc (%) | |
Without noise | 0.1 m | 0.03 | 0.004 | 98.73 | 2.41 | 0.186 | 96.40 |
1 m | 0.09 | 0.006 | 98.39 | 2.33 | 0.186 | 94.97 | |
3 m | 0.18 | 0.013 | 97.74 | 2.23 | 0.146 | 94.56 | |
5 m | 0.08 | 0.010 | 97.81 | 2.65 | 0.134 | 94.24 | |
With noise | 0.1 m | 0.96 | 0.046 | 97.20 | 2.30 | 0.123 | 95.15 |
1 m | 1.53 | 0.072 | 95.29 | 5.07 | 0.319 | 92.25 | |
3 m | 2.16 | 0.072 | 94.95 | 4.85 | 0.283 | 91.06 | |
5 m | 2.14 | 0.094 | 94.56 | 5.13 | 0.298 | 90.63 |