Front-End Factor Analysis for Speaker Verification

By Najim Dehak, Patrick Kenny, Réda Dehak, Pierre Dumouchel, Pierre Ouellet. 2011

3,573 citations recorded by the source corpus.

Subfields: Artificial Intelligence, Speech & Audio, Signal Processing

How researchers see this paper

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

In Artificial Intelligence

In Speech & Audio

In Signal Processing

Citations in the corpus

Jukebox: A Generative Model for Music (2020, methodology): “Ideas like clustering (Dehak et al., 2011), priming (Wang et al.”
An Unsupervised Autoregressive Model for Speech Representation Learning (2019, methodology): “For speaker verification, we follow the standard split of TIMIT, use the training set for training the universal background model, the i-vector extractor [29], a linear discriminant analysis (LDA) model.”
Exploring the Encoding Layer and Loss Function in End-to-End Speaker and Language Recognition System (2018, background): “In recent decades, the classical GMM i-vector approach and its variants have dominated multiple kinds of paralinguistic speech attribute recognition fields for its superior performance, simplicity and efficiency [3, 4].”
VoxCeleb2: Deep Speaker Recognition (2018, methodology): “Traditionally, the field of speaker recognition has been dominated by i-vectors [11], classified using techniques such as heavy-tailed PLDA [12] and GaussPLDA [13].”
Style Tokens: Unsupervised Style Modeling, Control and Transfer in End-to-End Speech Synthesis (2018, methodology): “Jauk (2017), for example, uses i -vectors (Dehak et al., 2011) and other acoustic features to cluster the training set and train models in different partitions.”
Generalized End-to-End Loss for Speaker Verification (2017, background): ““OK Google” [5, 6] In previous studies, i-vector based systems have been the dominating approach for both TD-SV and TI-SV applications [7].”
VoxCeleb: A Large-Scale Speaker Identification Dataset (2017, methodology): “The state of the art in more recent times involves both the use of joint factor analysis (JFA) based methods which model speaker and channel subspaces separately [9], and i-vectors which attempt to model both subspaces into a single compact, low-dimensional space [10].”
End-to-end text-dependent speaker verification (2015, methodology): “A simple scoring function is the cosine similarity between the speaker representation f ( X ) of an evaluation utterance X (see paragraph ”Training”) and the speaker model m spk (see paragraph ”Enrollment”): PLDA has been proposed as a more reﬁned, data-driven scoring approach.”
Unsupervised Learning of Disentangled and Interpretable Representations from Sequential Data (2017, methodology): “Learning to disentangle the mixed information from the surface representation is essential for a wide variety of speech applications: for example, noise robust speech recognition [41, 38, 37, 16], speaker verification [5], and voice conversion [40, 29, 24].”
Achieving Human Parity in Conversational Speech Recognition (2016, methodology): “Speaker adaptive modeling in our system is based on conditioning the network on an i-vector [49] characterization of each speaker [50, 51].”
Parallel training of Deep Neural Networks with Natural Gradient and Parameter Averaging (2014, methodology): “[30] Najim Dehak, Patrick Kenny, Réda Dehak, Pierre Dumouchel, and Pierre Ouellet, “Front-end factor analysis for speaker verification,” Audio, Speech, and Language Processing, IEEE Transactions on, vol.”
The CAPIO 2017 Conversational Speech Recognition System (2017, background): “, i-vectors [11] in case of speaker adaptation) for a better feature transformation [12, 13, 14].”
Disentangled Sequential Autoencoder (2018, methodology): “In formulas, We also include two baseline results from Hsu et al. (2017): one used the i-vector method (Dehak et al., 2011) for feature extraction, and the other one used µ 1 and µ 2 (analogous to µ z and µ f in our case) from a trained FHVAE model on Mel-scale ﬁlter bank (FBank) features.”
Deep Speaker: an End-to-End Neural Speaker Embedding System (2017, methodology): “Since gender labels are not available in all our datasets, the evaluation here is not split by gender, unlike the NIST speaker recognition evaluations (SREs) in [3].”
Phonetically-constrained PLDA modeling for text-dependent speaker verification with multiple short utterances (2013, methodology): “An i-vector is the compact representation of a variable-duration recording in a low-dimensional space called Total Variability space [4].”
Feature dimensionality reduction: a review (2022, methodology): “In this modeling [55], a new low-dimensional speaker- and channel-dependent space is defined using simple factor analysis.”
State-of-the-Art Speech Recognition Using Multi-Stream Self-Attention with Dilated 1D Convolutions (2019, methodology): “For the neural network acoustic models, we used the 40-dimensional higherresolution MFCCs appended with 100-dimensional i-vectors [33] and trained the models having the lattice alignments given by the SAT-ed GMMs as soft targets.”
rVAD: An Unsupervised Segment-Based Robust Voice Activity Detection Method (2019, methodology): “The TI-SV systems use 60dimension MFCCs as features and 600-dimension i-vector [73] together with probability linear discriminate analysis (PLDA) based scoring as the speaker verification back-end.”