Isabel Patiño

Head measurements are widely used in medicine, acoustics, and ergonomics. We focus on measurements required in acoustics to modify the Head Related Transfer Function. The modyfication of this function allows the placement of virtual sound sources in order to generate 3D virtual auditoriums. Such auditoriums enhance the experience of videoconferences and video games, among others. The measurements are automatically estimated on uncolored 3D point clouds, which are captured with an RGBD sensor Asus Xtion PRO LIVE and KinFu.

The required measurements are: head width, height, and depth, as well as the distance from the ear pits to the nose bridge and to the nape. In order to estimate the aforementioned dimensions, it is necessary to localize some feature points on the head such as: the nose tip, the ear pits, the top of the head, the chin, the nose bridge, the nape, the forehead, and the most prominent part on the back of the head.

Fig.1 Feature points and measurements. The head width is represented by w, h is the height, d represents the depth, f and b symbolize the distance from the ear pits to the nose bridge and to the nape, respectively. The features points are: 1. nose tip, 2. ear pit, 3. top of the head, 4. forehead, 5. chin, 6. nose bridge, 7. nape, 8. back of the head, 9. 3 cm in fron of the ear pits, used for the calculation of the head width.

The feature points are located on a complete (360°) point cloud of the subject, which is obtained while the subject rotates in front of the sensor. Since the 3D reconstruction quality is affected by the displacements of the subject when capturing the data, and in order to permit rotations in clockwise and counter-clockwise direction, the method uses two extra point clouds: right and left profile. These extra point clouds are employed to reduce the region of interest of the ear on the complete point cloud. Some results of the feature points location are shown below:

Fig.2 Visual results of the feature points location.