Getting More for Less: Using Weak Labels and AV-Mixup for Robust Audio-Visual Speaker Verification

TL;DR

This paper introduces AV-Mixup and a weakly supervised multitask learning approach to improve audio-visual speaker verification, achieving state-of-the-art results with reduced overfitting and enhanced speaker representations.

Contribution

It proposes a novel multimodal augmentation technique called AV-Mixup and demonstrates that weak labels in multitask learning can boost speaker verification performance.

Findings

Achieved state-of-the-art EER on VoxCeleb1-O/E/H datasets.

AV-Mixup reduces speaker overfitting during training.

Weak labels in multitask learning improve speaker representation quality.

Abstract

Distance Metric Learning (DML) has typically dominated the audio-visual speaker verification problem space, owing to strong performance in new and unseen classes. In our work, we explored multitask learning techniques to further enhance DML, and show that an auxiliary task with even weak labels can increase the quality of the learned speaker representation without increasing model complexity during inference. We also extend the Generalized End-to-End Loss (GE2E) to multimodal inputs and demonstrate that it can achieve competitive performance in an audio-visual space. Finally, we introduce AV-Mixup, a multimodal augmentation technique during training time that has shown to reduce speaker overfit. Our network achieves state of the art performance for speaker verification, reporting 0.244%, 0.252%, 0.441% Equal Error Rate (EER) on the VoxCeleb1-O/E/H test sets, which is to our knowledge, the best published results on VoxCeleb1-E and VoxCeleb1-H.