Improved Relation Networks for End-to-End Speaker Verification and Identification

TL;DR

This paper introduces improved relation networks combined with meta-learning for end-to-end speaker verification and few-shot speaker identification, demonstrating superior performance on multiple datasets.

Contribution

The paper proposes novel relation network architectures and training regimes for more effective speaker verification and identification, especially in few-shot scenarios.

Findings

Outperforms existing methods on VoxCeleb, SITW, and VCTK datasets.

Faster convergence with the new training regime.

Effective joint training of encoder and backend model.

Abstract

Speaker identification systems in a real-world scenario are tasked to identify a speaker amongst a set of enrolled speakers given just a few samples for each enrolled speaker. This paper demonstrates the effectiveness of meta-learning and relation networks for this use case. We propose improved relation networks for speaker verification and few-shot (unseen) speaker identification. The use of relation networks facilitates joint training of the frontend speaker encoder and the backend model. Inspired by the use of prototypical networks in speaker verification and to increase the discriminability of the speaker embeddings, we train the model to classify samples in the current episode amongst all speakers present in the training set. Furthermore, we propose a new training regime for faster model convergence by extracting more information from a given meta-learning episode with negligible extra computation. We evaluate the proposed techniques on VoxCeleb, SITW and VCTK datasets on the tasks of speaker verification and unseen speaker identification. The proposed approach outperforms the existing approaches consistently on both tasks.