Automatic player and ball position detection in beach volleyball video data

This research project, in cooperation with the Technical University of Munich and the German beach volleyball team, deals with the automatic detection of player and ball positions from video data. The results from the automatic tracking will help to improve the statistical video analysis in professional beach volleyball by providing more accurate and numerous data. Some of the techniques used are particle filters, kalman filters, histogram matching, fast radial symmetry transform, kernel based weighting and other modern methods of image processing and pattern recognition.

 

Externalization of sound in hearing-aids using Head Related Transfer Functions (HRTFs)

Spatial localization of sound sources by hearing aid users is often worse than that of normal hearing subjects, in particular when hearing impaired subjects are wearing closed canal hearing aids in both ears. Since the direct sound from the sound source to the tympanic membrane gets blocked by the ear mold and masked by the processed sound from the hearing aid, the spatial clues necessary to localize the source in space are partly lost. The aim of this research is to study the improvement in speech comprehension when externalizing sounds by filtering the signals with head related transfer functions (HRTFs).  Virtual spatial separation of the speech signal from other sound sources improves speech intelligibility on the one hand, and on the other hand it is more likeable for hearing aid users to actually hear the sound outside of the head, such as normal hearing subjects do, instead of an intracranial stereo image when using bilateral hearing aids.