VoxCeleb2: Deep Speaker Recognition

Chung, Joon Son; Nagrani, Arsha; Zisserman, Andrew

doi:10.21437/interspeech.2018-1929

Cited by 1,458 publications

(963 citation statements)

References 21 publications

Supporting

Mentioning

959

Contrasting

Unclassified

Order By: Relevance

“…We train our model on VoxCeleb2 [6], a large-scale audio-visual dataset of interviews obtained from unedited YouTube videos. The dataset consists of over a million utterances for 6,112 identities.…”

Section: Datasetmentioning

confidence: 99%

“…The motivation for doing so is simple: unlike earlier datasets such as TIMIT [5] that are carefully balanced for phonetic and dialectal coverage, more modern (and larger) datasets created from uncontrolled speech 'in the wild' are likely to contain a strong correlation between identity and linguistic content. For example, VoxCeleb2 [6] consists of interviews of famous celebrities from a wide variety of professions, whose speech can be closely tied to their occupation-the cricketer Adam Gilchrist says the word 'cricket' 17 times and 'president' 0 times; whereas the politician Nancy Pelosi says the word 'President' 88 times, 'Democrats' 19 times and 'cricket' 0 times. Consequently a model trained to represent identity may be incentivised to use linguistic content as a discriminative cue.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Disentangled Speech Embeddings Using Cross-Modal Self-Supervision

Nagrani

Chung

Albanie

et al. 2020

ICASSP 2020 - 2020 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP)

Self Cite

View full text Add to dashboard Cite

The objective of this paper is to learn representations of speaker identity without access to manually annotated data. To do so, we develop a self-supervised learning objective that exploits the natural cross-modal synchrony between faces and audio in video. The key idea behind our approach is to tease apart-without annotation-the representations of linguistic content and speaker identity. We construct a two-stream architecture which: (1) shares low-level features common to both representations; and (2) provides a natural mechanism for explicitly disentangling these factors, offering the potential for greater generalisation to novel combinations of content and identity and ultimately producing speaker identity representations that are more robust.We train our method on a large-scale audio-visual dataset of talking heads 'in the wild', and demonstrate its efficacy by evaluating the learned speaker representations for standard speaker recognition performance.

show abstract

Section: Datasetmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Disentangled Speech Embeddings Using Cross-Modal Self-Supervision

Nagrani

Chung

Albanie

et al. 2020

ICASSP 2020 - 2020 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP)

Self Cite

View full text Add to dashboard Cite

show abstract

“…To introduce more speaker variability during training, we additionally use data from VoxCeleb 1 and 2 [22,23]. We have kept all the speakers with more than 6 utterances each, resulting to a subset containing 6,490 speakers.…”

Section: Datasetsmentioning

confidence: 99%

A Memory Augmented Architecture for Continuous Speaker Identification in Meetings

Flemotomos

Dimitriadis

2020

ICASSP 2020 - 2020 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP)

View full text Add to dashboard Cite

We introduce and analyze a novel approach to the problem of speaker identification in multi-party recorded meetings. Given a speech segment and a set of available candidate profiles, we propose a novel data-driven way to model the distance relations between them, aiming at identifying the speaker label corresponding to that segment. To achieve this we employ a recurrent, memory-based architecture, since this class of neural networks has been shown to yield advanced performance in problems requiring relational reasoning. The proposed encoding of distance relations is shown to outperform traditional distance metrics, such as the cosine distance. Additional improvements are reported when the temporal continuity of the audio signals and the speaker changes is modeled in. In this paper, we have evaluated our method in two different tasks, i.e. scripted and real-world business meeting scenarios, where we report a relative reduction in speaker error rate of 39.28% and 51.84%, respectively, compared to the baseline.Index Termsspeaker identification, diarization, memory networks, meeting analysis * Nikolaos Flemotomos performed this work while at Microsoft. 1 A single speaker is present.

show abstract

“…Biometric authentication systems are generally based on either physiological biometrics such as fingerprints [1], face [2], [3], and voice [4], [5]), or behavioral biometrics such as touch [6] and gait [7], the latter category generally used for continuous and implicit authentication of users. These systems are mostly based on machine learning: a binary classifier is trained on the target user's data (positive class) and a subset of data from other users (negative class).…”

Section: Introductionmentioning

confidence: 99%

On the Resilience of Biometric Authentication Systems against Random Inputs

Zhao¹,

Asghar²,

Kâafar³

2020

Proceedings 2020 Network and Distributed System Security Symposium

View full text Add to dashboard Cite

We assess the security of machine learning based biometric authentication systems against an attacker who submits uniform random inputs, either as feature vectors or raw inputs, in order to find an accepting sample of a target user. The average false positive rate (FPR) of the system, i.e., the rate at which an impostor is incorrectly accepted as the legitimate user, may be interpreted as a measure of the success probability of such an attack. However, we show that the success rate is often higher than the FPR. In particular, for one reconstructed biometric system with an average FPR of 0.03, the success rate was as high as 0.78. This has implications for the security of the system, as an attacker with only the knowledge of the length of the feature space can impersonate the user with less than 2 attempts on average. We provide detailed analysis of why the attack is successful, and validate our results using four different biometric modalities and four different machine learning classifiers. Finally, we propose mitigation techniques that render such attacks ineffective, with little to no effect on the accuracy of the system.

show abstract

VoxCeleb2: Deep Speaker Recognition

Cited by 1,458 publications

References 21 publications

Disentangled Speech Embeddings Using Cross-Modal Self-Supervision

Disentangled Speech Embeddings Using Cross-Modal Self-Supervision

A Memory Augmented Architecture for Continuous Speaker Identification in Meetings

On the Resilience of Biometric Authentication Systems against Random Inputs

Contact Info

Product

Resources

About