Elisabeth van Pruissen (DTU MSc) will give a talk titled " Sound localization using a tetrahedral microphone configuration". The project is based on her MSc thesis work that was carried out at B&K. Please find the abstract below.
Sound localization using a tetrahedral microphone configuration
The interest of the automotive industry towards detection of buzz, squeak and rattle (BSR) noise inside a car cabin has grown over the years. As cars are getting more silent, these noises are becoming more apparent. The BSR’s play an important role in the perceived quality of the car by the
customer and should therefore be eliminated. Elimination is done in multiple stages of the production
process, but in this thesis the application for full vehicle tests is considered. A tetrahedral microphone
array is used to find the location of the BSR sources. Two acoustic source localization algorithms are
tested both in a simulation and in an experimental set-up (anechoic and inside a real car cabin). Different
variations of the steered-PHAT weighted generalized cross correlation are compared to variations of the
broadband Multiple Signal Classification (MUSIC) algorithm. Three new variations of the broadband
MUSIC algorithm are introduced, inspired by the GCC variations. These new variations outperform
the original broadband MUSIC algorithm, but they remain sensitive to reverberation and noise. This
showed to be especially problematic inside the car cabin. The GCC variations handle this reverberation
better and are able to detect a source inside a car cabin, even when the source is placed close to reflective
surfaces. The robustness towards the signal to noise ratio remains the limiting factor for the application
of BSR detection with the tetrahedral microphone array. For the three most often used full vehicle BSR
tests, the direct body excitation is the only one with noise levels that are low enough for BSR’s to be
detected with the investigated algorithms.