Photogrammetric computational methods can acquire precise personal head, external ear, and upper torso shapes using video captured with a mobile phone. We analyse the accuracy and repeatability of generating such 3D shape information. A known dummy head (Kemar) as well as a plastic dummy of a real person are considered. Resulting 3D data is compared to the manufacturer’s or actual 3D data. Shape errors are analysed. The HRTF estimates are compared. The impact of modeling errors is discussed. Photogrammetric determination of the 3D head-and-torso shape can have sufficient accuracy to allow computation of realistic personal head-related transfer function.
Authors: Aki Mäkivirta (Genelec), Matti Malinen (Kuava), Jaan Johansson (Kuava), Ville Saari (Genelec), Aapo Karjalainen (Kuava) and Poorang Vosough (Genelec)