A Comparison of Metric Learning Loss Functions for End-To-End Speaker Verification
This work addresses the lack of fair comparisons in metric learning for speaker verification, providing incremental improvements for researchers and practitioners in speech processing.
The paper systematically compares metric learning loss functions for speaker verification on the VoxCeleb dataset, finding that the additive angular margin loss outperforms others and learns more robust representations, while the network used is competitive with the x-vector baseline.
Despite the growing popularity of metric learning approaches, very little work has attempted to perform a fair comparison of these techniques for speaker verification. We try to fill this gap and compare several metric learning loss functions in a systematic manner on the VoxCeleb dataset. The first family of loss functions is derived from the cross entropy loss (usually used for supervised classification) and includes the congenerous cosine loss, the additive angular margin loss, and the center loss. The second family of loss functions focuses on the similarity between training samples and includes the contrastive loss and the triplet loss. We show that the additive angular margin loss function outperforms all other loss functions in the study, while learning more robust representations. Based on a combination of SincNet trainable features and the x-vector architecture, the network used in this paper brings us a step closer to a really-end-to-end speaker verification system, when combined with the additive angular margin loss, while still being competitive with the x-vector baseline. In the spirit of reproducible research, we also release open source Python code for reproducing our results, and share pretrained PyTorch models on torch.hub that can be used either directly or after fine-tuning.