Michel Dutat, Ivan Magrin-Chagnolleau, and Frédéric Bimbot.
Language Recognition using Time-Frequency
Principal Component Analysis and Acoustic Modeling.
Proceedings of ICSLP 2000, Beijing, China, October 2000.

Abstract: In this paper, we use a new speech parameterization
based on a principal component analysis applied to feature
parameters augmented by their time context.  This new parame-
terization is called time-frequency principal component (TFPC)
analysis.  We apply the new parameterization in the framework
of automatic language recognition. This new approach allows us
to improve the identification rate compared to the use of the
classical cepstral coefficients augmented by their delta-coefficients.