7th International Conference on Spoken Language Processing

September 16-20, 2002
Denver, Colorado, USA

Approaches to Language Identification Using Gaussian Mixture Models and Shifted Delta Cepstral Features

Pedro A. Torres-Carrasquillo (1), Elliot Singer (2), Mary A. Kohler (3), Richard J. Greene (2), Douglas A. Reynolds (2), J. R. Deller Jr. (1)

(1) Michigan State University, USA; (2) Massachusetts Institute of Technology, USA; (3) Department of Defense USA, USA

Published results indicate that automatic language identification (LID) systems that rely on multiple-language phone recognition and ngram language modeling produce the best performance in formal LID evaluations. By contrast, Gaussian mixture model (GMM) systems, which measure acoustic characteristics, are far more efficient computationally but have tended to provide inferior levels of performance. This paper describes two GMM-based approaches to language identi- fication that use shifted delta cepstra (SDC) feature vectors to achieve LID performance comparable to that of the best phone-based systems. The approaches include both acoustic scoring and a recently developed GMM tokenization system that is based on a variation of phonetic recognition and language modeling. System performance is evaluated on both the CallFriend and OGI corpora.

Full Paper

Bibliographic reference.  Torres-Carrasquillo, Pedro A. / Singer, Elliot / Kohler, Mary A. / Greene, Richard J. / Reynolds, Douglas A. / Deller Jr., J. R. (2002): "Approaches to language identification using Gaussian mixture models and shifted delta cepstral features", In ICSLP-2002, 89-92.