Voxceleb: Large-scale speaker verification in the wild

• Published in: Computer speech & language Vol. 60; p. 101027
• Main Authors: Nagrani, Arsha, Chung, Joon Son, Xie, Weidi, Zisserman, Andrew
• Format: Journal Article
• Language: English
• Published: Elsevier Ltd 01.03.2020
• Subjects: Convolutional neural network; Deep learning; Speaker identification; Speaker verification; Deep learning; Speaker identification; Speaker verification; Convolutional neural network
• ISSN: 0885-2308; 1095-8363
• DOI: 10.1016/j.csl.2019.101027

•We introduce the VoxCeleb dataset, the largest audio-visual dataset for speaker recognition containing over a million real world utterances from over 6000 speakers.•We develop a completely scalable, computer vision based pipeline to automatically create this dataset from open-source media.•We demonstrate that deep ResNet architectures trained on large datasets with NetVlad as an aggregation strategy achieve state of the art performance. The objective of this work is speaker recognition under noisy and unconstrained conditions. We make two key contributions. First, we introduce a very large-scale audio-visual dataset collected from open source media using a fully automated pipeline. Most existing datasets for speaker identification contain samples obtained under quite constrained conditions, and usually require manual annotations, hence are limited in size. We propose a pipeline based on computer vision techniques to create the dataset from open-source media. Our pipeline involves obtaining videos from YouTube; performing active speaker verification using a two-stream synchronization Convolutional Neural Network (CNN), and confirming the identity of the speaker using CNN based facial recognition. We use this pipeline to curate VoxCeleb which contains contains over a million ‘real-world’ utterances from over 6000 speakers. This is several times larger than any publicly available speaker recognition dataset. Second, we develop and compare different CNN architectures with various aggregation methods and training loss functions that can effectively recognise identities from voice under various conditions. The models trained on our dataset surpass the performance of previous works by a significant margin.