This paper describes Brno University of Technology (BUT) system for 2007 NIST Language recognition (LRE) evaluation. The system is a fusion of 4 acoustic and 9 phonotactic sub-systems. We have investigated several new topics such as discriminatively trained language models in phonotactic systems, and eigen-channel adaptation in model and feature domain in acoustic systems. We also point out the importance of calibration and fusion. All results are presented on NIST 2007 LRE data.
Bibliographic reference. Matějka, Pavel / Burget, Lukáš / Glembek, Ondřej / Schwarz, Petr / Hubeika, Valiantsina / Fapšo, Michal / Mikolov, Tomáš / Plchot, Oldřich / Černocký, Jan (2008): "BUT language recognition system for NIST 2007 evaluations", In INTERSPEECH-2008, 739-742.