Hi!

My name is Marijn Huijbregts. I'm a postdoc researcher at the Centre for Language and Speech Technology at Radboud University Nijmegen and I'm founder of a start-up company for audio search technology: Cross-Media Interaction (X-MI). This site contains some information about both jobs. If you want to make a more personal acquaintance, feel free to contact me!

My research

At the Centre for Language and Speech Technology at Radboud University Nijmegen, I'm currently working on speaker retrieval, but I'm also interested in Spoken Document Retrieval in general and key technologies for spoken document retrieval, such as Automatic Speech Recognition, Speech Activity Detection and Speaker Diarization. Here's a short overview of research that I have been involved in.

Speaker retrieval

Check out my speaker retrieval demo to learn more about this interesting research topic. Sorry, for now it's in Dutch only..

Spoken Document Retrieval

For my PhD research, I focussed on Spoken Document Retrieval (SDR). This technique makes it possible to search through audio or video the same way as it is possible to search through text documents (like search engines on the internet). Below I have drawn a rough diagram of the procedure.

First Automatic Speech Recognition (ASR) is used to translate the speech from audio or video files into text. The text and also the time in the video that it is pronounced are stored in a database. After this, when a user formulates a query, the system will search through the database and it will (hopefully) come up with some relevant audio or video fragments.

Automatic Speech Recognition, Speech Activity Detection and Speaker Diarization

Automatic speech recognition is a key technology in spoken document retrieval, but for most spoken document retrieval tasks, ASR on its own is not enough. Two other key techniques are Speech Activity Detection (SAD) and Speaker Diarization. SAD removes all audible non-speech from a recording and speaker diarization (the task of: "who speaks when?") is used to optimize ASR. Check out my publications (in particular my thesis) for more information on these three technologies.

Publications

Presentations

Sorry, this list is not up-to-date. I'll try to update it as soon as possible.

Oral presentations

Poster presentations

X-MI

Cross-Media Interaction

Together with Roeland Ordelman from the University of Twente, I have founded Cross-Media Interaction (X-MI). At X-MI we apply our research to real-live tasks. For more information about my X-MI work, please visit the X-MI site.

SHOUT

Large Vocabulary Continuous Speech Recognition Toolkit

During my PhD research at the University of Twente, I have developed the open source speech recognition toolkit "SHoUT". SHoUT is the Dutch acronym for "Speech Recognition Research at the University of Twente". Sounds better than the English acronym "SRRatUT" don't you think?

Check out my Sourceforge page for more information!

Hobby

So that's it for my professional life. But wait... There's more!
I also have a personal life! That part of the site is in Dutch though.