Improving Speaker Identification For Shared Devices By Adapting Embeddings To Speaker Subsets
2021 · Zhenning Tan, Yuguang Yang, Eunjung Han, et al.
Abstract
Speaker identification typically involves three stages. First, a front-end speaker embedding model is trained to embed utterance and speaker profiles. Second, a scoring function is applied between a runtime utterance and each speaker profile. Finally, the speaker is identified using nearest neighbor according to the scoring metric. To better distinguish speakers sharing a device within the same household, we propose a household-adapted nonlinear mapping to a low dimensional space to complement the global scoring metric. The combined scoring function is optimized on labeled or pseudo-labeled speaker utterances. With input dropout, the proposed scoring model reduces EER by 45-71% in simulated households with 2 to 7 hard-to-discriminate speakers per household. On real-world internal data, the EER reduction is 49.2%. From t-SNE visualization, we also show that clusters formed by household-adapted speaker embeddings are more compact and uniformly distributed, compared to clusters formed by
Authors
(none)
Tags
Stats
Related papers
- Graph-based Multi-view Fusion And Local Adaptation: Mitigating Within-household Confusability For Speaker Identification (2022)2.26
- Unified Hypersphere Embedding For Speaker Recognition (2018)0.00
- Graph-based Label Propagation For Semi-supervised Speaker Identification (2021)8.09
- Adapting End-to-end Neural Speaker Verification To New Languages And Recording Conditions With Adversarial Training (2018)9.59
- Few-shot Speaker Identification Using Lightweight Prototypical Network With Feature Grouping And Interaction (2023)9.03
- Deep Speaker Embeddings For Far-field Speaker Recognition On Short Utterances (2020)11.29
- Analyzing Speaker Verification Embedding Extractors And Back-ends Under Language And Channel Mismatch (2022)0.00
- End-to-end Recurrent Denoising Autoencoder Embeddings For Speaker Identification (2020)6.34