Efficient Auralization by Grouping Directions and Modeling HRTFs Using Wavelets
Júlio César, Bárbara Torres Gómez de Cádiz, Roberto A. Tenenbaum, Mariane R. Petraglia, Greek Symbols · 2007
In the last ten years, it can be observed an increas ing number of immersive audio systems. This growing is due mainly to new technologies trying to simulate the human s ensation to be immersed in a real ambient. An example is the inclusion of audio tracks in DVDs recorded with dummy h eads, trying to recreate the recording space in the reproduction. However, systems of acoustical virtual real ity, also called auralization, require a very high computat ional complexity to reproduce the 3-D characteristics of the actu al sound. One of the best ways to reduce this computational complexity is to model, in an efficient and realistic way, thetransfer functions related to the human head, the HRTFs, using wavelets and sparse filters, which is reported in this p aper. Furthermore, since the HRTFs are not much sensitive to subtle changes in directions (depending of course on the f requency range), a new scheme to group directions in a judicious way is described, with simulation results, to red uce even more the computational complexity. The main idea is that, since the selectivity of the wavelet filters is high, it is possible to use a common set of coefficients for the same band of all HRTFs of a region. An efficient auralization schem e, exploring the similarity among the model coefficients is presented, and some different grouping settling are tested . The main conclusion is that the adopted techniques improve noticeably the auralization with an error lower than 30%.