Accuracy of three dimensional head model acquired by optical scanning and its effect on the head-related transfer function

Personalized three-dimensional (3D) head models have important applications in head-related transfer function (HRTF) and virtual auditory display (VAD). Optical scanning is a common method to obtain 3D head models, but the deviations of different optical scanning methods and their effects on HRTF are not clear. In this paper, the self-designed 3D head model served as a reference model M_O (baseline), and the 3D-printed fabricated model is used as the scanning object. Three representative optical scanning devices (handheld red laser scanner HandySCAN 300, handheld white light-emitting diode (LED) scanner Reeyee Pro 2X, and infrared laser iPhone 12 cell phone) are selected to scan and obtain the corresponding models M_H, M_R, and M_I. To evaluate the surface accuracy of the newly acquired models, the color-coded mapping of the graphical deviations between the models and the root mean square (RMS) values of the deviation distances are compared. Further, the HRTF spectral errors between models are compared to analyze the effect of model accuracy on HRTF data. The results show that the graphical deviations are mainly concentrated in the partial details of the auricle, and the HRTF universal errors mainly appear in the high frequencies. In comparison, model M_H is the closest to M_O, with surface errors basically within ±0.40 mm, and HRTF spectral errors exceeding 3.0 dB occur only in the frequency band above 8 kHz, and such errors are observed only in a few spatial directions. The surface errors of the models M_R and M_I compared to M_O reach about −0.90 mm to 1.90 mm and −1.20 mm to 3.20 mm, respectively, and the HRTF spectral errors above 8 kHz exceed 3.0 dB in most directions. The study provides a reference basis for the selection of 3D head scanning modalities.