Semi-Supervised Contrastive Learning with Generalized Contrastive Loss and Its Application to Speaker Recognition

TL;DR

This paper presents a semi-supervised contrastive learning framework using a generalized contrastive loss that unifies supervised and unsupervised learning, applied effectively to speaker verification tasks.

Contribution

It introduces a generalized contrastive loss that seamlessly integrates supervised and unsupervised learning for speaker recognition.

Findings

GCL enables learning speaker embeddings in supervised, semi-supervised, and unsupervised modes.

The framework improves speaker verification performance on VoxCeleb dataset.

Unified loss function simplifies semi-supervised learning implementation.

Abstract

This paper introduces a semi-supervised contrastive learning framework and its application to text-independent speaker verification. The proposed framework employs generalized contrastive loss (GCL). GCL unifies losses from two different learning frameworks, supervised metric learning and unsupervised contrastive learning, and thus it naturally determines the loss for semi-supervised learning. In experiments, we applied the proposed framework to text-independent speaker verification on the VoxCeleb dataset. We demonstrate that GCL enables the learning of speaker embeddings in three manners, supervised learning, semi-supervised learning, and unsupervised learning, without any changes in the definition of the loss function.