Evaluation of the accuracy of photogrammetry for head-related transfer functions acquisition using numerical methods
* Presenting author
Abstract:
The use of photogrammetry to create meshes suitable for deterministic numerical methods is thought to be a promising approach for individual Head-Related Transfer Functions (HRTFs) acquisition. The validity of this HRTFs extraction technique is discussed in this paper. Photogrammetric methods are employed to reconstruct the shape of a Neumann KU100 dummy head from a set of pictures captured with a consumer-grade digital camera. The accuracy of the resulting surface information is analyzed by comparing it with a precise reconstruction obtained through laser scanning. Numerical HRTFs computed on these geometries are compared to a set of acoustic HRTFs measured on the same dummy head. The validity of the computed transfer functions is further assessed by means of a virtual perceptual model focusing on elevation perception on the sagittal plane. The results indicate that the adopted method is capable of providing accurate HRTFs in terms of binaural cues related to horizontal plane localization. However, the use of HRTFs derived from photogrammetry tends to hinder the elevation perception which relates to perceived realism in a complex virtual scene. This suggests that the photogrammetric method used in this paper lacks the precision needed for the acquisition of accurate HRTFs.