The SONICOM HRTF Dataset has been created within the SONICOM project and includes different types of data from several subjects (currently 200, but further measurements are ongoing). More details about how the dataset has been measured can be found in our Journal of Audio Engineering Society paper here.
For every subject HRIRs are sampled at 96 kHz and 24 bits, but lower sample rates (44.1 and 48 kHz) are also included. For each sample rate, the following versions of the HRTFs can be downloaded in SOFA format:
- Raw
- Windowed
- Windowed and Free-field compensated
- Windowed and Free-field compensated with a minimum phase filter (i.e. the compensation filter - not the HRTF - is minimum phase)
All the above are also provided with the ITDs removed from the signals and stored as metadata in the SOFA files. Some of the versions above are also released in the .3dti format, in order for them to be directly usable with the 3DTune-In Toolkit.
HpTFs (Sennheiser HD650) for every subject are also included in .mat format at each of the sample rates listed above.
The 3D models of ears, head and torso are available in .stl format, with and without watertight post-processing. Image data is available upon request (one image every 5° of horizontal rotation across the whole head) as .tif (depth), along with .txt (gravity) files. RGB image data is also available upon request (contact o.turvey22@imperial.ac.uk).
HRTF and 3D Scan data can be explored and downloaded at the following link:
Anthropometric pinna parameters were collected for 93 subjects, following the measurement specifications of the HUTUBS dataset [1]. The parameters are in the file SONICOM_anthropometries.csv, where each row corresponds to a single ear for a given subject. The column "Ear" indicates whether the entry refers to the left or right ear. The IDs in the column "SONICOM ID" correspond to the subject IDs used in the dataset and appear twice across rows to account for both ears of each subject. Distance parameters (d1 to d10) are provided in centimeters, and angle parameters (θ1 and θ2) in degrees.
Parameters d1 to d10 were manually measured on the watertight 3D models using Blender’s "Measure" tool, with careful selection of the corresponding points on each pinna model. The angle parameters θ1 and θ2 were extracted using the same semi-automatic procedure as in the HUTUBS dataset. In this process, four characteristic points are manually placed on the 3D pinna model, and a Python script executed within Blender automatically computes the angle measurements.
(Manual measurements by Matteo Castagna; Supervisors: Davide Fantini and Federico Avanzini; University of Milan)
[1] Brinkmann, F., Dinakaran, M., Pelzer, R., Grosche, P., Voss, D., & Weinzierl, S. (2019). A cross-evaluated database of measured and simulated HRTFs including 3D head meshes, anthropometric features, and headphone impulse responses. Journal of the Audio Engineering Society, 67(9), 705-718.
The HRTFs of the KEMAR dummy head are also available at the link above. These include:
- Large and Small ears (Large Ears: KB0065/KB0066, Small Ears: KB0060/KB0061)
- Measurements with the same Knowles microphones used for the other participants - in this case, we used the GRAS 2CC ear simulators in order to be able to insert the microphones in the canals. (Ear Simulator Product Number: RA0045)
- Measurements with GRAS microphones, 1/8" when using the 2CC ear simulator/canal (with microphone in the eardrum position) and 1/4" without the ear simulator/canal (with microphone at the entrance of the ear canal)
- For the Knowles microphones version with Large ears, we also measured the full HRTF with the head turned 45° and 90° relative to the shoulders and torso.
The following additional files are also available:
- ChinRest.zip -contains the .stl file of the custom chin rest, as well as a few pictures of it mounted on the chair structure
- TurnTable.zip -contains the documentation for the custom turntable, including the CAD model (URL), the Arduino code for controlling the rotation, as well as a Quick Start guide
- Pictures.zip -some pictures of the setup in the Turret Lab room
The iPhone mirror bracket is based on this design.
The SONICOM HRTF Dataset is publicly available for non-commercial purposes.
We are currently measuring subjects - if you are interested in participating, please book your slot at this link: https://axdteam.simplybook.it/v2/