Jon Gudnason

Post-Doc Research Assistant
B.Sc. M.Sc. Ph.D.

I graduated from the University of Iceland in June 2000 with a M.Sc. in Signal Processing. My thesis was on System Identification of Speech with Recurrent Neural Networks. Since then I have worked with Mike Brookes and Patrick Naylor on speaker recognition. One of our research interests is the voice production model and how to extract glottal source features. Another area of interest is further development and use of these features in statistical pattern classifiers.

Research Interests
My research interests lie in signal processing and pattern recognition particularly when applied to speech. I'm particularly interested in development of robust feature sets for speech applications such as speaker recognition. The research projects I have been involved with are the following

Novel Features for Speaker Recognition (VERIPHON):
We use glottal source features extracted from the speech waveform to improve the performance of speech verification and identification systems.

Radar Identification of Targets with Hidden Markov Models at Imperial College (RITHMIC)
This project aims to identify moving targets using high range resolution (HRR) radar signatures.

Detection of Unconvertible Voice-mail Messages for Spinvox
Spinvox provides a voice-mail to text-message conversion service for telephone users. The speech quality of such messages is highly variable. This project aims to automatically detect messages that are of so low quality that they are unconvertible to text.

List of publications

Software (Speaker Identification with YOHO)

Interesting Links