Study Of Head-related Transfer Function Of KEMAR Based On Multiple Datasets

Head-related transfer functions(HRTFs) are the core of virtual auditory display, which reflect the interaction of free-field sound waves and human anatomical structures(especially the head and ears) when transmitting from sound source to human ears. Due to individuality of human anatomical structures, HRTFs are individual-dependent. Based on the different populations, many human-subject HRTF databases have been established. If the existing databases can be combined, then the research of individual HRTF can be promoted. However, different databases used different measuring methods, equipments, and post-processing methods. Therefore, the combination of HRTF databases should be on the basis of reasonable assessment of different HRTF measurements and effective pre-processing method.This study uses KEMAR(Knowles Electronics Mannequin for Acoustics Research) as the research object. Using subjective listening experiments and objective indicator analysis, differences between the five KEMAR HRTF databases are assessed systematically. Objective analysis shows that using diffusion field equalization method as the pre-processing method not only can effectively extract the spatial location information in HRTFs, but also can reduce the difference between different HRTF measurements. After diffusion field equalization, the spectral difference among different HRTF measurements reduces to about 3 d B. Listening experiment results show that although there are certain timbre differences among some databases, the five KEMAR HRTF databases are equivalent in terms of localization.Because pinna is the main anatomical origin of HRTF’s individulity, this paper also applies principal component analysis(PCA) to study the influence of pinnae on HRTF individulity. Analysis shows that the impact of pinnae on HRTF is mainly at some specific frequency bands and azimuths. In details, pinnae differences cause prominent HRTF difference above 6 k Hz and at the ipsilateral side of studied pinnae.On the basis of the five KEMAR HRTF database, this paper evaluates the effect of different measurements on HRTFs by using subjective and objective methods, and validates the diffusion field equalization method as an effective pre-processing method. Moreover, this paper also studies the effect of typical pinnae on HRTF individuality. Results of this study can not only deepen our understanding of HRTF characteristics represented by KEMAR, but also provide a reference for the combination of human-subject HRTF databases in the future.