Estimating the total variability space using sparse probabilistic principal component analysis

Authors:
Zhenchun Lei;Jihua Ye
Affiliations:
School of Computer and Information Engineering, Jiangxi Normal University, Nanchang, China;School of Computer and Information Engineering, Jiangxi Normal University, Nanchang, China
Venue:
CCBR'12 Proceedings of the 7th Chinese conference on Biometric Recognition
Year:
2012

Citing 4
Cited 0

EM algorithms for PCA and SPCA

NIPS '97 Proceedings of the 1997 conference on Advances in neural information processing systems 10
Mixtures of probabilistic principal component analyzers

Neural Computation
PCA versus LDA

IEEE Transactions on Pattern Analysis and Machine Intelligence
Joint Factor Analysis Versus Eigenchannels in Speaker Recognition

IEEE Transactions on Audio, Speech, and Language Processing

Quantified Score

Hi-index	0.00

Visualization

Abstract

In this paper, we introduce a new method to estimate the total variability space using sparse probabilistic Principal Component Analysis (PCA) with the Baum-Welch statistics for speaker verification. In conventional method, probabilistic PCA is used, which is a probabilistic formulation for PCA. Recently some methods improve interpretability by sparse representation through adding an L1 regularizer. We introduce a Laplacian prior to each element in the transformation matrix, since Laplacian prior is equivalent to L1 regularization. Variational inference is used and we can drive the EM algorithm formulas for estimating the space with the statistics. After WCCN, the cosine similarity scoring is used for decision. The experiments have been run on the NIST SRE 2008 data, and the results show that the performance can be improved 10.2% for female and is comparable for male.