An Efficient DSP Network for the Real Time Auralisation of Complex Urban Scenarios
Master Thesis of Andrew, Henry
Even with modern hardware, many virtual seen es are aehallenge to auralise in real time. Part of the problem is that eurrent renderers struggle to manage and apply the neeessary operations to the large number of paths needed for eomplex scenes. Outdoor urban seenes are partieularly diffieult due to many sourees, whieh may be fast moving, and lots of diffraeted and refiected paths. A novel rendering module is presented in this thesis which implements several recentconcepts not found in other renderers, and is built with an efficient DSP structure. Key features include: using IIR filters in place of more conventional FIR filters to simulate air absorption and diffraction effects, path clustering to limit the number of HRTF directions, and interpolation from multi read cursor variable delay lines to implement Doppler shifting. This thesis was implemented in the VA framework, and uses several libraries developed at the Institute of Technical Acoustics at RWTH Aaehen university. During the development of the rendering module, several algorithms for IIR filter design were tested. The Burg algorithm was chosen as the design method for the IIR path filters due to its fast eomputation time and high quality filter design.