The Phonexia VoxCeleb Speaker Recognition Challenge 2021 System Description

TL;DR

This paper details the Phonexia system for the VoxCeleb Speaker Recognition Challenge 2021, utilizing unsupervised learning, clustering, and score fusion to improve speaker verification accuracy.

Contribution

It introduces a novel unsupervised speaker verification approach combining contrastive learning, clustering, and ensemble scoring, inspired by prior winning methods.

Findings

Effective unsupervised embedding extraction via contrastive learning

Iterative clustering improves pseudo-label quality

Score fusion enhances verification performance

Abstract

We describe the Phonexia submission for the VoxCeleb Speaker Recognition Challenge 2021 (VoxSRC-21) in the unsupervised speaker verification track. Our solution was very similar to IDLab's winning submission for VoxSRC-20. An embedding extractor was bootstrapped using momentum contrastive learning, with input augmentations as the only source of supervision. This was followed by several iterations of clustering to assign pseudo-speaker labels that were then used for supervised embedding extractor training. Finally, a score fusion was done, by averaging the zt-normalized cosine scores of five different embedding extractors. We briefly also describe unsuccessful solutions involving i-vectors instead of DNN embeddings and PLDA instead of cosine scoring.