Aki Mäkivirta, Matti Malinen, Ville Saari, Juha Holm, Jaan Johansson, Aapo Karjalainen, Poorang Vosough, Juha Urhonen
The presentation discusses a method for obtaining accurate HRTF data for a high number of sound arrival azimuth and elevation angles, using a sequence of images of head and torso extracted from a video captured while circulating a person. High-quality mobile phone cameras are sufficient for capturing the video. A detailed 3D model scaled to correct dimensions of the head and upper torso is created using stereophotogrammetric methods, with mesh density for modelling of external ears. After this, acoustic fields are calculated numerically with a full-wave method to capture detailed acoustic phenomena. Acoustic fields are calculated for over 800 different orientations for the audio to approach the head. Then, the HRTFs are formed and finally the data is compiled into a downloadable SOFA file. Initial evaluation of the quality of the resulting HRTF set is presented.