In defence of metric learning for speaker recognition
share
The objective of this paper is 'open-set' speaker recognition of unseen speakers, where ideal embeddings should be able to condense information into a compact utterance-level representation that has small intra-class (same speaker) and large inter-class (different speakers) distance. A popular belief in speaker recognition is that networks trained with classification objectives outperform metric learning methods. In this paper, we present an extensive evaluation of most recent loss functions for speaker recognition on the VoxCeleb dataset. We demonstrate that even the vanilla triplet loss shows competitive performance compared to classification-based losses, and those trained with our angular metric learning objective outperform state-of-the-art methods.
READ FULL TEXT