Probabilistic Linear Discriminant Analysis for Inferences About Identity

By Simon J. D. Prince, James H. Elder. 2007

1,155 citations recorded by the source corpus.

Subfields: Computer Vision and Pattern Recognition

How researchers see this paper

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

In Computer Vision and Pattern Recognition

Citations in the corpus

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.”
Neural Voice Cloning with a Few Samples (2018, methodology): “We use the probabilistic linear discriminant analysis (PLDA) for scoring the similarity between the two encodings [Prince and Elder, 2007, Snyder et al., 2016].”
Deep Speaker: an End-to-End Neural Speaker Embedding System (2017, methodology): “Traditionally, i-vectors have been used to model inter-speaker variability [3]. i-vector-based speaker recognition models perform clas-siﬁcation using cosine similarity between i-vectors or more advanced techniques such as PLDA [20], heavy-tailed PLDA [21], and Gauss-PLDA [5].”
ASVspoof 2019: Future Horizons in Spoofed and Fake Audio Detection (2019, methodology): “The common ASV system uses x-vector speaker embed-dings [14] together with a probabilistic linear discriminant analysis (PLDA) [19] backend.”
The NIST 2014 Speaker Recognition i-vector Machine Learning Challenge (2014, background): “The current state-of-theart scoring function, called Probabilistic Linear Discriminant Analysis (PLDA) [5] [6], requires a within-class (WC) matrix, characterizing how i-vectors from a single speaker vary, and 10 .”
Spoofing and countermeasures for speaker verification: A survey (2015, neutral): “, 2012), a factor analysis technique originally developed for face recognition (Prince and Elder, 2007), is the most popular approach.”
Full-covariance UBM and heavy-tailed PLDA in i-vector speaker verification (2011, methodology): “We use a Probabilistic Linear Discriminant Analysis Model (PLDA) [3] that provides a probabilistic framework applicable to fixed-length input vectors.”
Generalized I-vector Representation with Phonetic Tokenizations and Tandem Features for both Text Independent and Text Dependent Speaker Verification (2015, background): “Finally, PLDA modeling and score level fusion is presented in Sec 2.3.”
A Divide-and-Conquer Approach for Language Identification Based on Recurrent Neural Networks (2016, methodology): “Here, the PLDA (Probabilistic Linear Discriminant Analysis) technique [27], which is also commonly used for speaker verification [24, 25] or gender identification [28], was used.”
Bayesian reasoning and machine learning (2012, neutral): “The model can be trained using either a constrained form of the direct method, or EM as described in [228].”
Analysis of i-vector Length Normalization in Speaker Recognition Systems (2011, methodology): “Examples following these assumptions are the Probabilistic Linear Discriminant Analysis (PLDA) model proposed in [3] and the two-covariance (2-cov) model introduced in [4].”
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].”
Multi-Directional Multi-Level Dual-Cross Patterns for Robust Face Recognition (2014, methodology): “Component Analysis (WPCA; an unsupervised learning algorithm) and Probabilistic Linear Discriminant Analysis (PLDA; a supervised learning algorithm) [33], [34].”
A Comprehensive Study on Center Loss for Deep Face Recognition (2019, background): “Belhumeur et al. (1997), Wang and Tang (2004), Prince and Elder (2007) decomposed the face representations into several components to facilitate the subsequent recognition.”
ASVspoof 2019: A large-scale public database of synthesized, converted and replayed speech (2019, methodology): “The original Kaldi recipe was modiﬁed to include PLDA adaptation using disjoint, bona ﬁde, in-domain data.”
Voices Obscured in Complex Environmental Settings (VOICES) corpus (2018, methodology): “The model used is a Universal Background Model (UBM) identity vector (i-vector) based system [13, 14], with a probabilistic linear discriminant analysis (PLDA) [15] as back-end classiﬁer.”