Related Experiment Video
Updated: Nov 23, 2025
07:34
Perceptual and Category Processing of the Uncanny Valley Hypothesis' Dimension of Human Likeness: Some Methodological Issues
Published on: June 3, 2013
17.7K
Head-related transfer function recommendation based on perceptual similarities and anthropometric features
Robert Pelzer1, Manoj Dinakaran1, Fabian Brinkmann1
1Audio Communication Group, Technical University of Berlin, Einsteinufer 17c, D-10587, Germany.
The Journal of the Acoustical Society of America
|December 31, 2020
Summary
Individualizing head-related transfer functions (HRTFs) using anthropometric features (AFs) improves binaural audio quality. Machine learning models predict perceived differences, recommending HRTFs that enhance localization and reduce coloration.
Area of Science:
- Acoustics
- Perception
- Virtual Reality Audio
Background:
- Head-related transfer functions (HRTFs) are crucial for realistic binaural audio.
- Individualizing HRTFs enhances localization accuracy and reduces coloration.
- Anthropometric features (AFs) offer a non-invasive method for HRTF individualization.
Purpose of the Study:
- To directly investigate the perceptual impact of anthropometric features on HRTF individualization.
- To develop a machine learning model predicting perceived differences in HRTFs based on AFs.
- To identify key AFs influencing perceived coloration and localization errors.
Main Methods:
- A listening test was conducted where subjects compared their individual HRTFs with non-individual ones.
- Machine learning models were trained to predict perceived HRTF differences using ratios of anthropometric features.
- The relevance of specific AFs for predicting perceptual errors was analyzed.
Main Results:
- The developed machine learning models accurately predicted perceived differences in HRTFs.
- Recommended HRTFs significantly improved upon generic or randomly selected HRTFs.
- Key anthropometric features influencing perceived coloration and localization errors were identified.
Conclusions:
- Anthropometric features can effectively predict perceptual differences in HRTFs, enabling personalized audio experiences.
- The developed models offer a practical approach to HRTF individualization, improving binaural application quality.
- The models and findings are publicly available under a free cultural license.
