@inproceedings{f590a21e7fc44c67be031aad557f72a5,
title = "Unifying probabilistic linear discriminant analysis variants in biometric authentication",
abstract = "Probabilistic linear discriminant analysis (PLDA) is commonly used in biometric authentication. We review three PLDA variants - standard, simplified and two-covariance - and show how they are related. These clarifications are important because the variants were introduced in literature without argumenting their benefits. We analyse their predictive power, covariance structure and provide scalable algorithms for straightforward implementation of all the three variants. Experiments involve state-of-the-art speaker verification with i-vector features.",
keywords = "i-vectors, PLDA, speaker and face recognition",
author = "Aleksandr Sizov and Lee, {Kong Aik} and Tomi Kinnunen",
year = "2014",
doi = "10.1007/978-3-662-44415-3_47",
language = "English",
isbn = "9783662444146",
series = "Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)",
publisher = "Springer Verlag",
pages = "464--475",
booktitle = "Structural, Syntactic, and Statistical Pattern Recognition - Joint IAPR International Workshop, S+SSPR 2014, Proceedings",
address = "Germany",
note = "Joint IAPR International Workshop on Structural, Syntactic, and Statistical Pattern Recognition, S+SSPR 2014 ; Conference date: 20-08-2014 Through 22-08-2014",
}