An efficient wavelet-based HRTF model for auralization

In this paper, a new method for modeling Head-Related Transfer Functions (HRTFs) is presented. This model is based on the decomposition of the impulse response of the HRTFs through wavelet transforms, whose coefficients can be obtained either by an adaptive filtering algorithm or by an analytical formulation – both of which are presented in this paper. The structure has a set of sparse filters, following the wavelet transform, which implements the desired impulse response. The influence of the wavelet basis and structure on the modeled coefficients are investigated. Through an analysis of the behavior of the coefficients and by using a criterion based on the energy contribution of each sparse filter coefficient, it is shown how the model can be reduced without introducing considerable error in the magnitude and phase frequency responses. The detailed structure for obtaining the model and the criteria adopted to reduce the coefficients are presented. As a result of the proposed technique, the reduced model has approximately 30% of the number of coefficients of the original HRTF, which represents an important reduction in the computational cost for implementing an auralization system.