I-vector extraction and Probabilistic Linear Discriminant Analysis (PLDA) has become the state-of-the-art configuration for speaker verification. Recently, Gaussian-PLDA has been improved by a preliminary length normalization of i-vectors. This normalization, known to increase the Gaussianity of the i-vector distribution, also improves performance of systems based on standard Linear Discriminant Analysis (LDA) and ”two-covariance model” scoring. But this technique follows a standardization of the i-vectors (centering and whitening ivectors based on the first and second order moments of the development data). We propose in this paper two techniques of normalization based on total, between- and within-speaker variance spectra. These ”spectral” techniques both normalize the i-vectors length for Gaussianity, but the first adapts the ivectors representation to a speaker recognition system based on LDA and two-covariance scoring when the second adapts it to a Gaussian-PLDA model. Significant performance improvements are demonstrated on the male and female telephone portion of NIST SRE 2010.
Index Terms: i-vectors, probabilistic linear discriminant analysis, speaker recognition.
Cite as: Bousquet, P.-M., Larcher, A., Matrouf, D., Bonastre, J.-F., Plchot, O. (2012) Variance-spectra based normalization for i-vector standard and probabilistic linear discriminant analysis. Proc. The Speaker and Language Recognition Workshop (Odyssey 2012), 157-164
@inproceedings{bousquet12_odyssey,
author={Pierre-Michel Bousquet and Anthony Larcher and Driss Matrouf and Jean-François Bonastre and Oldřich Plchot},
title={{Variance-spectra based normalization for i-vector standard and probabilistic linear discriminant analysis}},
year=2012,
booktitle={Proc. The Speaker and Language Recognition Workshop (Odyssey 2012)},
pages={157--164}
}