Adaptive Margin Circle Loss for Speaker Verification
This work addresses speaker verification for security and biometric applications, presenting an incremental improvement over existing angular loss methods.
The paper tackles speaker verification by proposing an adaptive margin circle loss, which improves angular discrimination and achieves state-of-the-art results with 1.31% EER on Voxceleb1 and 2.13% on SITW core-core.
Deep-Neural-Network (DNN) based speaker verification sys-tems use the angular softmax loss with margin penalties toenhance the intra-class compactness of speaker embeddings,which achieved remarkable performance. In this paper, we pro-pose a novel angular loss function called adaptive margin cir-cle loss for speaker verification. The stage-based margin andchunk-based margin are applied to improve the angular discrim-ination of circle loss on the training set. The analysis on gradi-ents shows that, compared with the previous angular loss likeAdditive Margin Softmax(Am-Softmax), circle loss has flexi-ble optimization and definite convergence status. Experimentsare carried out on the Voxceleb and SITW. By applying adap-tive margin circle loss, our best system achieves 1.31%EER onVoxceleb1 and 2.13% on SITW core-core.