Bayesian Speaker Verification with Heavy-Tailed Priors

By Patrick Kenny. 2010

801 citations recorded by the source corpus.

Subfields: Probabilistic & Bayesian Methods, Speech & Audio

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Probabilistic & Bayesian Methods

In Speech & Audio

Citations in the corpus

CHiME-6 Challenge: Tackling Multispeaker Speech Recognition for Unsegmented Recordings (2020, methodology): “A probabilistic linear discriminant analysis (PLDA) model [45] is trained on the CHiME-6 dataset.”
Exploring the Encoding Layer and Loss Function in End-to-End Speaker and Language Recognition System (2018, methodology): “• Decision generator, includes logistic regression (Lo-gReg), support vector machine (SVM), and neural network for closed-set identiﬁcation, cosine similarity or probabilistic linear discriminant analysis (PLDA) [17, 18] for open-set veriﬁcation.”
End-to-end text-dependent speaker verification (2015, methodology): “For example, GMM-UBM [7] or i-vector models does not directly optimize a verification problem; the PLDA [5] model is not followed a refinement of the i-vector extraction; or long contextual features may be ignored by frame-based GMM-UBM models [7].”
Text-dependent speaker recognition using PLDA with uncertainty propagation (2013, background): “Joint-Factor Analysis and its successor, the combination of i-vectors and Probabilistic Linear Discriminant Analysis (PLDA) have become the dominant approaches in the field, as the NIST and other contests can verify, [1], [2], [3].”
Speaker Recognition for Multi-speaker Conversations Using X-vectors (2019, methodology): “A probabilistic linear discriminant analysis (PLDA) [3] classifier is used to compare i-vectors, and enable same-or-different speaker decisions [4, 5].”
Comparison of Speaker Recognition Approaches for Real Applications (2011, methodology): “It is worth noting that LDA WCCN and Gaussian PLDA training and scoring have been performed by using L2 normalized i-vectors.”
Improving speaker recognition performance in the domain adaptation challenge using deep neural networks (2014, positive): “Current speaker recognition systems model i-vectors [1] with variants of Probabilistic Linear Discriminant Analysis (PLDA) [2, 3, 4, 5, 6, 7].”
A Review of Speaker Diarization: Recent Advances with Deep Learning (2021, methodology): “In the following studies [130, 131], a method applying”
Non-parallel voice conversion using i-vector PLDA: towards unifying speaker verification and transformation (2017, methodology): “Besides the original formulation in [7], there are other flavors to PLDA, including simplified [8] and two-covariance [9] PLDA models (for a comparison and scalable implementations of all the three, refer to [20]).”
ASVspoof 2015: the first automatic speaker verification spoofing and countermeasures challenge (2015, methodology): “In order to confirm vulnerabilities to spoofing, experiments were conducted using the challenge database and a state-of-theart Probabilistic Linear Discriminant Analysis (PLDA) [29, 30] ASV system.”
ASVspoof 2019: A large-scale public database of synthesized, converted and replayed speech (2019, methodology): “The second offline step is to learn a subspace in the ivector space that best discriminates speakers and acquire a speaker factor extractor via a probabilistic linear discriminant (PLDA) model [47][48].”
Bayesian HMM clustering of x-vector sequences (VBx) in speaker diarization: theory, implementation and analysis on standard tasks (2020, methodology): “discrimination between speakers, the speaker- (or HMM state-) specific distributions are derived from a PLDA [21] model pre-trained on a large number of speaker-labeled x-vectors.”
Deep Speaker Embeddings for Short-Duration Speaker Verification (2017, methodology): “PLDA explicitly models the speaker and channel separately and thus performs channel compensation as part of verification [5].”
AutoSpeech: Neural Architecture Search for Speaker Recognition (2020, positive): “Earlier speaker recognition systems used spectral features (e.g., MFCCs) as the speaker embedding [24, 25], but these systems have been superseded by systems based on i-vectors [26, 27, 28].”
X-Vectors: Robust DNN Embeddings for Speaker Recognition (2018, methodology): “A probabilistic linear discriminant analysis (PLDA) [3] classifier is used to compare i-vectors, and enable same-or-different speaker decisions [4, 5, 6].”
Self-Attentive Speaker Embeddings for Text-Independent Speaker Verification (2018, background): “Over the years, the combination of i-vectors [1] and probabilistic linear discriminant analysis (PLDA) [2] has been the dominant approach for text-independent SV tasks [3, 4, 5].”
Time delay deep neural network-based universal background models for speaker recognition (2015, methodology): “In this paradigm, a universal background model (UBM) is used to collect sufficient statistics for i-vector extraction, and a probabilistic linear discriminant analysis (PLDA) backend computes a similarity score between i-vectors [2, 3, 4, 5, 6, 7].”
Front-End Factor Analysis For Speaker Verification (2018, methodology): “After having successfully been applied in face recognition, PLDA [2] was introduced by Kenny in speaker verification [14] to model in a generative way the distribution of the ivectors.”