Generation of individualized head-related transfer functions in the upper hemisphere using the Parametric Notch-Peak HRTF model

In order to realize three-dimensional sound image localization and acoustic VR, methods for generating head-related transfer functions (HRTFs) for individual listeners have been actively studied. It is reported that the two notches (N1 and N2) and the two peaks (P1 and Pd) in the HRTF play important role in sound image localization in the median plane. We proposed an HRTF model (Parametric Notch-Peak model) that can continuously change the parameters corresponding to individual differences by utilizing the knowledge on the N1, N2, P1 and Pd. In the present paper, we developed seven PNP models in the upper median plane. Then, the individualized HRTFs for arbitrary directions in the upper median plane were generated using the seven PNP models. Finally, the individualized HRTFs for arbitrary directions in the upper hemisphere were generated by adding the interaural difference cues to the individualized HRTFs in the upper median plane. Furthermore, we have developed an application software that generates individualized HRTFs based on the proposed method. Results of the sound image localization tests showed that the individualized HRTFs provide accurate sound image localization in the horizontal plane and the upper median plane.