Reconstruction head-related transfer function from individualized

Far-field Head-Related Transfer Functions (HRTFs) vary with source direction, frequency and individual. Correspondingly, the dimensionality of a full set of HRTFs is huge and measurement or calculation of individualized HRTFs with high directional resolution could be difficult. A method to construct HRTFs with high directional resolution from a small set of directional measurements or calculations is proposed in present paper. Based on tensor decomposition of HRTFs, far-field HRTFs are decomposed into a combination of directional modes, frequency modes and a few of individual modes. The individual-independent matrices of directional modes and frequency modes are derived by statistical analysis on a known base-line HRTF database. For arbitrary new individual, the individual mode can be estimated from measured or calculated HRTFs at a small set of directions, then the HRTFs with high directional resolution are reconstructed. Calculation from two base-line HRTF databases indicates that 11 individual modes account for more than 98% individual-related energy variation of HRTFs; and HRTFs with high directional resolution can be reconstructed from about 30 directional measurements or calculation. A psychoacoustic experiment validates the analysis. The present method is applicable to simplify individualized HRTF measurement or calculation.