The Ins And Outs Of Speaker Recognition: Lessons From Voxsrc 2020

Abstract

The VoxCeleb Speaker Recognition Challenge (VoxSRC) at Interspeech 2020 offers a challenging evaluation for speaker recognition systems, which includes celebrities playing different parts in movies. The goal of this work is robust speaker recognition of utterances recorded in these challenging environments. We utilise variants of the popular ResNet architecture for speaker recognition and perform extensive experiments using a range of loss functions and training parameters. To this end, we optimise an efficient training framework that allows powerful models to be trained with limited time and resources. Our trained models demonstrate improvements over most existing works with lighter models and a simple pipeline. The paper shares the lessons learned from our participation in the challenge.

Stats

• citations37
• S2 citations—
• github stars0
• HF likes0
• heat score11.85
• arxiv keykwon2020the

Related papers

• Voxsrc 2019: The First Voxceleb Speaker Recognition Challenge (2019)0.00
• Voxsrc 2020: The Second Voxceleb Speaker Recognition Challenge (2020)0.00
• Voxceleb2: Deep Speaker Recognition (2018)23.96
• Clova Baseline System For The Voxceleb Speaker Recognition Challenge 2020 (2020)0.00
• The Phonexia Voxceleb Speaker Recognition Challenge 2021 System Description (2021)0.00
• Delving Into Voxceleb: Environment Invariant Speaker Recognition (2019)10.35
• The Xx205 System For The Voxceleb Speaker Recognition Challenge 2020 (2020)0.00
• Voxceleb: A Large-scale Speaker Identification Dataset (2017)23.55