Tensor modeling and interpolation for distance-dependent head-related transfer function

Head-related transfer functions (HRTFs) are multi-dimensional functions of source position, frequency, and anthropometric parameters. In this paper, a tensor is adopted to represent HRTFs dependent on elevation, azimuth, and frequency. The core tensor dependent on a source distance is extracted to capture most of the variations in the original tensor space. Interpolation between the low dimensional core tensors can obtain the desired HRTFs at an arbitrary distance according to the measurements. The tensor model compresses the original high dimensional data and realizes simple interpolation with high accuracy. Simulation results demonstrate the performance of the tensor modeling and interpolation for distance-dependent HRTFs.