Efficient State-Space Rational Interpolation of HRTFs

摘要

The paper addresses the task of spatially interpolating the head related transfer function (HRTF) for correct updating of binaural 3D sound synthesis in response to head-movements in telepresence environments. Using HRTF data, we create proper matrix transfer functions or equivalently appropriate state-space realizations. We base our method on the factorization of a block Loewner matrix into a product of generalized observability and controllability matrices. We recollect certain properties to be satisfied by the Loewner matrix, and use it to construct a minimal state-space realization of an interpolating matrix transfer function. Compared to existing interpolation methods, this novel approach proved to be outperforming and capable of insuring high-fidelity reconstructions of HRTFs.