Optimizing sound localization clustering using spherical Haar wavelets through multi-resolution analysis

A virtual sonic environment is a critical component of virtual reality applications that enhance the interactive experience to improve the perception of a virtual world. The fundamental feature of a virtual sonic environment is three-dimensional sound. In three-dimensional sound presentation, sound sources are filtered with a head related transfer function pair, one for each ear, in order to add the perception of three-dimensional localization to sounds. These filtering operations can overburden computer resources as the number of sounds increase. This dissertation proposes an innovative algorithm to control the expendability of computational resources and sound localization resolution. This control is achieved by applying the multi-resolution analysis theory to human sound localization through the utilization of spherical wavelets. Spherical wavelets define sets of spherical triangular regions with different sizes. The algorithm proposed here utilizes the spherical triangular regions to cluster sounds according to their position. The number of clusters bounds the computational resources required to process the three-dimensional sounds. This method constitutes a powerful practical and mathematical tool that has not been applied before to synthesize sound localization. The proposed method's functionality is tested through perceptual experiments. The results demonstrate the effectiveness of controlling sound localization resolution.