Predicting Personalized Head Related Transfer Functions using Acoustic Scattering Neural Networks

Javeri N, Dutta PB, Sunder K, Jain K. Predicting Personalized Head Related Transfer Functions using Acoustic Scattering Neural Networks. In: AES Convention 153. Audio Engineering Society; 2022. Paper 53. Available from: https://aes.org/publications/elibrary-page/?id=21937

@inproceedings{Javeri2022_21937,
  author = {Javeri, Nikhil and Dutta, Prabal Bijoy and Sunder, Kaushik and Jain, Kapil},
  title = {{Predicting Personalized Head Related Transfer Functions using Acoustic Scattering Neural Networks}},
  booktitle = {AES Convention 153},
  note = {Paper 53},
  year = {2022},
  month = oct,
  publisher = {Audio Engineering Society},
  url = {https://aes.org/publications/elibrary-page/?id=21937}
}

TY  - CPAPER
TI  - Predicting Personalized Head Related Transfer Functions using Acoustic Scattering Neural Networks
AU  - Javeri, Nikhil
AU  - Dutta, Prabal Bijoy
AU  - Sunder, Kaushik
AU  - Jain, Kapil
T2  - AES Convention 153
M1  - Paper 53
PY  - 2022
DA  - 2022/10/06
UR  - https://aes.org/publications/elibrary-page/?id=21937
PB  - Audio Engineering Society
LA  - en
AB  - The quality of spatial audio production has become critical to deliver a truly immersive sound experience in the recent past. Binaural audio is one of the most convenient formats to deliver accurate spatial audio over headphones. Personalized Head-related Transfer Functions (HRTFs) are an integral component of binaural audio that determines the quality of the spatial audio experience. In this paper, we describe a novel technique to predict personalized HRTFs based on 2D images or a video capture. The state-of-the-art 3D reconstruction techniques were developed for generic objects and thus do not perform well with complex structures such as an ear. We propose a novel 3D reconstruction algorithm that is modeled taking into account the geometry of the ear. The 3D output is then fed to a Acoustic Scattering Neural Network (ASNN) designed on the principles of Boundary Element Method (BEM) that outputs personalized HRTFs. The personalized HRTFs predicted are then compared both objectively with the measured HRTFs. We discuss the results, limitations, and the caveats necessary for an accurate modeling of personalized HRTFs.
ER  -