Comparison of speaker recognition approaches for real applications

Cumani, Sandro; Batzu, Pier Domenico; Colibro, Daniele; Vair, Claudio; Laface, Pietro; Vasilakakis, Vasileios

doi:10.21437/interspeech.2011-64

Cited by 51 publications

(28 citation statements)

References 7 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…We create scores for all fine-tuned systems using adaptive s-normalization [18] with an imposter cohort size of 100. The imposter cohort consists of the average of the length-normalized utterance-based embeddings of each training speaker.…”

Section: Large Margin Fine-tuningmentioning

confidence: 99%

The IDLAB VoxCeleb Speaker Recognition Challenge 2020 System Description

Thienpondt,

Desplanques,

Demuynck

2020

Preprint

View full text Add to dashboard Cite

In this technical report we describe the IDLAB top-scoring submissions for the VoxCeleb Speaker Recognition Challenge 2020 (VoxSRC-20) in the supervised and unsupervised speaker verification tracks. For the supervised verification tracks we trained 6 state-of-the-art ECAPA-TDNN systems and 4 Resnet34 based systems with architectural variations. On all models we apply a large margin fine-tuning strategy, which enables the training procedure to use higher margin penalties by using longer training utterances. In addition, we use qualityaware score calibration which introduces quality metrics in the calibration system to generate more consistent scores across varying levels of utterance conditions. A fusion of all systems with both enhancements applied led to the first place on the open and closed supervised verification tracks. The unsupervised system is trained through contrastive learning. Subsequent pseudo-label generation by iterative clustering of the training embeddings allows the use of supervised techniques. This procedure led to the winning submission on the unsupervised track, and its performance is closing in on supervised training.

show abstract

Section: Large Margin Fine-tuningmentioning

confidence: 99%

The IDLAB VoxCeleb Speaker Recognition Challenge 2020 System Description

Thienpondt,

Desplanques,

Demuynck

2020

Preprint

View full text Add to dashboard Cite

show abstract

“…In contrast to the methods reviewed thus far, other methods attempt to overcome domain mismatch given existing embeddings. Score Normalization (SN) takes a single trial (enrollment, test recordings, and corresponding matching score) and normalizes the score according to the score distributions of the enrollment/test recordings with respect to an imposter dataset [17]. PLDA Adaptation (PA) [18] uses an unlabeled adaptation dataset to modify the PLDA model used to score the samples.…”

Section: Related Workmentioning

confidence: 99%

Siamese X-Vector Reconstruction for Domain Adapted Speaker Recognition

Rozenberg¹,

Aronowitz

Hoory

2020

Interspeech 2020

View full text Add to dashboard Cite

With the rise of voice-activated applications, the need for speaker recognition is rapidly increasing. The x-vector, an embedding approach based on a deep neural network (DNN), is considered the state-of-the-art when proper end-to-end training is not feasible. However, the accuracy significantly decreases when recording conditions (noise, sample rate, etc.) are mismatched, either between the x-vector training data and the target data or between enrollment and test data. We introduce the Siamese x-vector Reconstruction (SVR) for domain adaptation. We reconstruct the embedding of a higher quality signal from a lower quality counterpart using a lean auxiliary Siamese DNN. We evaluate our method on several mismatch scenarios and demonstrate significant improvement over the baseline.

show abstract

“…The most common way of calculating a similarity between a pair of face embeddings is to simply use a cosine similarity: sim(•, •) ≡ cos (•, •). Nevertheless, we have empirically observed that normalizing the cosine similarities with Adaptive-Symmetric NORMalization (as-norm) [32] allows to largely avoid the false positive verification errors (which are the most costly ones according to the challenges' metrics). Therefore, we employ as-norm for score normalization in our final pipeline (the IJC-B dataset [33] containing about 3500 identities is used to calculate the reference cohort).…”

Section: Calculation Of Similarity Between Face Embeddingsmentioning

confidence: 99%

Automatic Quality Assessment for Audio-Visual Verification Systems. The LOVe Submission to NIST SRE Challenge 2019

Antipov¹,

Gengembre²,

Blouch³

et al. 2020

Interspeech 2020

View full text Add to dashboard Cite

Fusion of scores is a cornerstone of multimodal biometric systems composed of independent unimodal parts. In this work, we focus on quality-dependent fusion for speaker-face verification. To this end, we propose a universal model which can be trained for automatic quality assessment of both face and speaker modalities. This model estimates the quality of representations produced by unimodal systems which are then used to enhance the score-level fusion of speaker and face verification modules. We demonstrate the improvements brought by this quality-dependent fusion on the recent NIST SRE19 Audio-Visual Challenge dataset.

show abstract

Comparison of speaker recognition approaches for real applications

Cited by 51 publications

References 7 publications

The IDLAB VoxCeleb Speaker Recognition Challenge 2020 System Description

The IDLAB VoxCeleb Speaker Recognition Challenge 2020 System Description

Siamese X-Vector Reconstruction for Domain Adapted Speaker Recognition

Automatic Quality Assessment for Audio-Visual Verification Systems. The LOVe Submission to NIST SRE Challenge 2019

Contact Info

Product

Resources

About