Short news

Dr. Paja

On Thursday 9th April, Pavel "Paja" Matejka successfully defended is PhD thesis "PHONOTACTIC AND ACOUSTIC LANGUAGE RECOGNITION". Paja was registered at the Faculty of Electrical Engineering and Communication (FEEC BUT) and defended at the Institute of Radioelectronics (the original place of the Speech@FIT group, before it moved to the Faculty of Information Technology in 2002).

He obtained very nice reviews from Doug Reynolds (MIT Lincoln Lab) and Petr Pollak (Czech Technical University in Prague) and the committee was happy with his presentation accessible even to non-speech guys.

Niko Brummer and Albert Strasheim in Brno

From 19.3.-3.4., Niko Brummer and Albert Strasheim from Agnitio were visiting Speech@FIT. Agnitio and BUT are preparing common submission to NIST 2009 language recognition evaluation, and preparation (and good calibration!) of the system was exactly what the two guys worked on with Speech@FITters (especially Ondra Glembek). Niko also did 2 excellent seminars.

Some photos from improvised farewel party:
Niko's seminar #1:

Marijn Huijbregts helping Speech@FITters with NIST RT 2009 evaluations

From 16 to 26 Februray 2009, Marijn Huijbregts from University of Twente visited the group to help us with diarization for NIST Rich Transcription 2009 evaluation. Marijn previously visited ICSI in Berkeley and has lots of experience in diarization. As in several past years, Speech@FIT participates in the AMIDA team building an LVCSR system for meeting data.

Some links:
The seminar he had at FIT (including video):

First round of MOBIO recordings finished at BUT

Last week, we have finished the BUT portion of recording bi-modal speaker authentication data for the MOBIO database: Chris McCool from IDIAP was here and helped start the recordings, then Speech@FIT's data manager Kamil Chalupnicek took over. More than 30 volunteers were recorded and they will be recorded another 5 times to study session variability. Similar amounts of volunteers are being recorded at other MOBIO partners.

New DCGM coffee machine

On Oct 14th, new coffee machine was delivered to the Department of Computer Graphics and Multimedia of FIT, which is Speech@FIT's base. It is "Rancilio Set of Silvia Coffee Machine, Rocky No Doser Coffee Grinder on Inox Base"

everything needs to be done by hand, but the espresso is ... simply excellent ! Some guys consider it much more important than victories in NIST evaluations or big grants, and have proposed 14th October as DCGM state holiday.

Speech@FIT received an EOARD grant for research in language recognition

The group was granted a 80kUS$ grant from US Air Force European Office of Aerospace Research & Development (EOARD). The goal of the project is to improve the capacity of language recognition systems to handle rare languages using radio broadcast data. For more information, see

Gokhan Ilk from University of Ankara visiting Speech@FIT

On 25.9.2008, Gokhan Ilk from University of Ankara visited the group in the framework of Erasmus/Socrates staff exchange. Gokhan gave a very nice seminar on DCT Interpolation of Speech Signals, which made us revise some basic truths about DCT. More on

Franta suffering in the cave (last time!)

Franta Grezl, Speech@FIT researcher, spends 22-24/9/2008 in Metro company of Roma [Italy] fine-tuning the demonstrators for the CareTaker final technical review. The computers are placed in a "cave" at the Termini station with some 35 degrees of temperature and 100% humidity. This time, our Roma partners promised to arrange for air-conditioners, though. Cross fingers to Franta and success of the CareTaker review taking place on 24/9. More info on the project is at:

Talk of Liz Shriberg at TSD

the talk was excellent and Honza is happy that he helped to bring Liz to Brno. See for summary of her materials on prosody. And ... we had a nice trip to Moravsky Kras, Jedovnice, and Krtiny ;-)

Web-based demo for Language Identification

The demo for on-line testing of language identification is now available at
It can detect Arabic, English, Farsi, French, German, Hindu, Japanese, Korean, Mandarin, Spanish, Tamil, Vietnamese, Czech, Polish and Russian. Standard WAV format is supported. Just try it out !