Table 6 Comparison with current state-of-the-art results on AP17-OLR in terms of EER and Cavg
From: Masked multi-center angular margin loss for language recognition
System | Test-all | Test-3 s | Test-1 s | |||
|---|---|---|---|---|---|---|
EER | Cavg | EER | Cavg | EER | Cavg | |
i-vecotr + PLDA [47] | 5.96 | 5.86 | - | - | 17.51 | 17.46 |
TDNN [47] | 11.31 | 10.34 | - | - | 14.04 | 12.82 |
LDA_HVS [48] | 3.19 | 3.30 | 3.79 | 3.77 | 8.47 | 8.82 |
AFs_ivector + AFs_xvector + AFs_TDNN [49] | 1.92 | 1.84 | - | - | - | - |
Multi-head attention [50] | 5.65 | - | - | - | - | - |
Wav2vec [51] | 3.47 | 3.10 | - | - | 12.02 | 11.58 |
MMAM(ours) | 1.35 | 1.23 | 1.48 | 1.47 | 7.78 | 7.28 |
Absolute improvement | 0.57 | 0.61 | 2.31 | 2.3 | 0.69 | 1.54 |
Relative improvement | 29.6% | 32.7% | 60.9% | 61% | 8.1% | 17.4% |