On the Relevance of Age, Sex, and Ethnicity for Listener Anthropometry, HRTFs, and Modeled Auditory Perception

Jain S, Brinkmann F, Haubrichs B, Moeller R, Weinzierl S. On the Relevance of Age, Sex, and Ethnicity for Listener Anthropometry, HRTFs, and Modeled Auditory Perception. J Audio Eng Soc. 2026;74(9):594-611. doi:10.17743/jaes.2026.0282

@article{Jain2026_23436,
  author = {Jain, Sanket and Brinkmann, Fabian and Haubrichs, Bjarne and Moeller, Rainer and Weinzierl, Stefan},
  title = {{On the Relevance of Age, Sex, and Ethnicity for Listener Anthropometry, HRTFs, and Modeled Auditory Perception}},
  journal = {Journal of the Audio Engineering Society},
  volume = {74},
  number = {9},
  pages = {594--611},
  year = {2026},
  month = sep,
  publisher = {Audio Engineering Society},
  doi = {10.17743/jaes.2026.0282},
  url = {https://doi.org/10.17743/jaes.2026.0282}
}

TY  - JOUR
TI  - On the Relevance of Age, Sex, and Ethnicity for Listener Anthropometry, HRTFs, and Modeled Auditory Perception
AU  - Jain, Sanket
AU  - Brinkmann, Fabian
AU  - Haubrichs, Bjarne
AU  - Moeller, Rainer
AU  - Weinzierl, Stefan
T2  - Journal of the Audio Engineering Society
J2  - J. Audio Eng. Soc.
VL  - 74
IS  - 9
SP  - 594
EP  - 611
PY  - 2026
DA  - 2026/09/16
DO  - 10.17743/jaes.2026.0282
UR  - https://doi.org/10.17743/jaes.2026.0282
PB  - Audio Engineering Society
LA  - en
AB  - Head-related transfer functions (HRTFs) are essential for spatial audio reproduction via headphones, and personalized HRTFs provide more natural audio rendering for a specific listener. Because measuring HRTFs for each individual is usually too costly, most personalization approaches including HRTF prediction via deep learning are based on HRTF databases whose representativeness for a larger population has hardly been investigated. To investigate how biases in HRTF databases might affect personalization, 3D head meshes and anthropometric measures were acquired, and HRTFs of 256 individuals were numerically calculated by equally sampling from 11 age groups, two sexes, and two ethnicities in populations in Sindelfingen, Germany, and Beijing, China. A comparison with existing data showed that the acquired anthropometric features cover a wider range of values, suggesting that the data here are more diverse. Statistical analyses using general linear models confirmed that ear and head size are significantly influenced by age, sex, and ethnicity, in accordance with previous studies. These factors also significantly affect coloration and median and horizontal plane localization errors modeled for pairs of nonindividual HRTFs. Although the effects were small, this suggests that HRTF personalization approaches could benefit from considering age, sex, and ethnicity.
ER  -