Exploring Binary Classification Loss For Speaker Verification

Bing Han, Zhengyang Chen, Yanmin Qian

The mismatch between close-set training and open-set testing usually leads to significant performance degradation for speaker verification task. For existing loss functions, metric…