Curriculum learning for self-supervised speaker verification

TL;DR

This paper introduces curriculum learning strategies for self-supervised speaker verification, progressively increasing data complexity to improve speaker representations, achieving state-of-the-art results on VoxCeleb1.

Contribution

It proposes two novel curriculum learning methods within a self-supervised framework to enhance speaker verification without labels.

Findings

Achieved 4.47% EER with single-phase training

Further improved to 1.84% EER with fine-tuning

Demonstrated effectiveness of curriculum strategies on VoxCeleb1

Abstract

The goal of this paper is to train effective self-supervised speaker representations without identity labels. We propose two curriculum learning strategies within a self-supervised learning framework. The first strategy aims to gradually increase the number of speakers in the training phase by enlarging the used portion of the train dataset. The second strategy applies various data augmentations to more utterances within a mini-batch as the training proceeds. A range of experiments conducted using the DINO self-supervised framework on the VoxCeleb1 evaluation protocol demonstrates the effectiveness of our proposed curriculum learning strategies. We report a competitive equal error rate of 4.47% with a single-phase training, and we also demonstrate that the performance further improves to 1.84% by fine-tuning on a small labelled dataset.