Feature Adaptation Based on Gaussian Posteriors

Author

Kozat, Suleyman S. ; Visweswariah, Karthik ; Gopinath, Ramesh

Author_Institution

IBM Thomas J. Watson Res. Center, Yorktown Heights, NY

Volume

1

fYear

2006

fDate

14-19 May 2006

Abstract

In this paper we consider the use of non-linear methods for feature adaptation to reduce the mismatch between test and training conditions. The non-linearity is introduced by using the posteriors of a set of Gaussians to (softly) partition the observation space for feature adaptation. The modeling framework used is based on the fMPE models (D. Povey et al., 2005) applied to FMLLR matrices directly. However, the parameters are estimated to maximize the likelihood of the test data. We observe a relative gain of 14% on top of FMLLR, which was a 42% relative gain over the baseline

Keywords

Gaussian processes; matrix algebra; maximum likelihood estimation; regression analysis; speech recognition; Gaussian posteriors; feature adaptation; maximum likelihood linear regression matrices; nonlinear methods; speech recognition; Acoustic testing; Adaptation model; Loudspeakers; Maximum likelihood linear regression; Parameter estimation; Piecewise linear techniques; Probability; Spatial databases; Speech recognition; System testing;

fLanguage

English

Publisher

ieee

Conference_Titel

Acoustics, Speech and Signal Processing, 2006. ICASSP 2006 Proceedings. 2006 IEEE International Conference on

Conference_Location

Toulouse

ISSN

1520-6149

Print_ISBN

1-4244-0469-X

Type

conf

DOI

10.1109/ICASSP.2006.1659997

Filename

1659997