Facebook Twitter LinkedIn You TUBE


Pavel Matějka

Czech Republic

Mr. Pavel Matějka


Future Forces Forum Future of Cyber Conference - Live Hacking Zone (FCC) 2018 BIOMETRICS CONFERENCE Future Forces Exhibition 2018


Brno University of Technology


Presentation

Math and algorithms behind speaker verification - from Gaussian models to neural nets 

The talk introduces the speaker recognition and its technological evolution from Gaussian Mixture model to Deep Neural Networks. The basic idea of each technology will be presented in the graphical form for better imagination (math behind is too complicated for 20 minutes talk). The motivation between each evolution step will be supported by the results. The results are presented on the data from evaluations organized by National Institute of Standard and Technology (NIST) from USA.


Curriculum Vitae

Pavel Matejka (Ing. [MS]. Brno University of Technology, 2001, PhD. Brno University of Technology, 2009) is a senior researcher at BUT Speech@FIT research group of the Department of Computer Graphics and Multimedia, Faculty of Information Technology, BUT. He was a member of the Anthropic speech processing group at the Oregon Graduate Institute of Science and Technology, USA, from 10/2002 till 6/2003 and member of research staff at Raytheon BBN from 2013 till 2014. He participated in the European Commission's projects M4, AMI, AMIDA, MOBIO, in the language identification projects sponsored by the US Air-Force European Office of Aerospace Research and Development (EOARD), several projects from the Czech Ministry of Defense and Interiors and also projects from US DARPA and IARPA (RATS, BABEL, MATERIAL). He took part in the NIST language recognition evaluation since 2003 and in the NIST speaker recognition evaluations since 2006, where BUT has continuously excellent results.

Pavel is author or co-author of more than 30 papers that were published in professional journals and reviewed at international conferences. He is a member of IEEE and ISCA. His research interests include robust speaker verification, language identification, speech recognition, namely phone recognition based on novel feature extractions (temporal patterns) and neural networks. 

© 2010 – 2018 Future Forces FORUM Prague, All rights reserved.