Improving Speaker Identification for Shared Devices by Adapting Embeddings to Speaker Subsets

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 global embeddings before adaptation.

arXiv commentSubmitted to ASRU 2021

References in corpus (9)

• Federated Learning: Strategies for Improving Communication Efficiency
• VoxCeleb2: Deep Speaker Recognition
• VoxCeleb: a large-scale speaker identification dataset
• Unsupervised Data Augmentation for Consistency Training
• ECAPA-TDNN: Emphasized Channel Attention, Propagation and Aggregation in TDNN Based Speaker Verification
• Deep Speaker: an End-to-End Neural Speaker Embedding System
• In defence of metric learning for speaker recognition
• Clova Baseline System for the VoxCeleb Speaker Recognition Challenge 2020
• Graph-based Label Propagation for Semi-Supervised Speaker Identification

Cited by in corpus (2)

• Improving fairness in speaker verification via Group-adapted Fusion Network
• Graph-based Multi-View Fusion and Local Adaptation: Mitigating Within-Household Confusability for Speaker Identification