Ir. Henk Brouckxon

Biography

Henk Brouckxon received the Burgerlijk Elektrotechnisch Ingenieur (Master of Applied Sciences in Electronics and Information Processing) from the Vrije Universiteit Brussel (VUB) in 2002. He is a PhD student at the Laboratory of Digital Speech and Audio Processing in the VUB Department of Electronics and Informatics (ETRO-DSSP)where his research is focused on digital signal processing algorithms for intelligibility evaluation and improvement of spoken messages presented in an adverse acoustical environment (e.g. public addressing in a noisy train station). He also assisted in the practical exercises classes for the Digital Signal processing class by prof. Werner Verhelst and supervised Masters theses on psycho-acoustics based audio requantisation, speech analysis algorithms and 3D audio auralisation (3D surround sound). His research interests include speech, speech intelligibility, psycho-acoustics, digital audio processing techniques, microphone arrays (localisation and beamforming) and surround sound auralisation.

Research Interest

When humans converse in a noisy environment, they employ a type of speech that is especially easy to understand and that is referred to as "clear speech". Studies have shown that comprehension of syllables is 13 % higher for clear speech than for normal conversational speech. In speech that is delivered to the listener through technological means, e.g. via telephones or announcement systems, a normal mode of speech is usually used as the speaker is not aware of the (possibly adverse) listening conditions. In this project, we study signal processing techniques which can be used to improve the intelligibility of speech in adverse acoustical environments. Ideally, we would like to find simple acoustic and phonetic differences between normal and clear speech, and develop a speech modification technology that can be used to automatically convert spoken messages from the normal mode to the clear mode. In this thesis we investigate digital signal processing algorithms that can be used to improve the acquisition and presentation of speech in noisy environments, with the goal of evaluating and improving its intelligibility. The research is primarily based on applications for public addressing systems, but can be applied more widely to any speech-based communication system that may be used in adverse acoustical environments. Public Addressing ($PA$) systems are used to present spoken informational and safety messages in large venues. The intelligibility of these systems can however be severely degraded by noise and reverberation. In this project, we therefore investigate algorithms that detect possible intelligibility problems and improve the public addressing system's intelligibility in noisy environments.