A Smoothing Method of Head-Related Transfer Functions Based on Reconstruction from Wavelet Transform Modulus Maxima

In the research of spatial hearing and implementation of virtual auditory space, it is important to accurately model the latent acoustical clues in HRTFs(Head-Related Transfer Functions) or HRIRs(Head-Related Impulse Responses) related to certain position of sound source. As an essential preprocessing step, this work introduced a new smoothing means based on trous algorithm with translation-invariant and reconstruction from modulus maxima, and managed to carry through adaptive non-linear approximation in the field of wavelet transformation. The simulation results show that, under the uniform threshold, the performance of the new way is averagely 8.3dB better than that of traditional PCA(Principal Component Analysis) method, and 2.4dB than that of wavelet method using Mallat algorithm.