Deep Learning for Synthesis of Head-Related Transfer Functions

Bharitkar SG. Deep Learning for Synthesis of Head-Related Transfer Functions. In: AES Convention 146. Audio Engineering Society; 2019. Paper 10161. Available from: https://aes.org/publications/elibrary-page/?id=20294

@inproceedings{Bharitkar2019_20294,
  author = {Bharitkar, Sunil G.},
  title = {{Deep Learning for Synthesis of Head-Related Transfer Functions}},
  booktitle = {AES Convention 146},
  note = {Paper 10161},
  year = {2019},
  month = mar,
  publisher = {Audio Engineering Society},
  url = {https://aes.org/publications/elibrary-page/?id=20294}
}

TY  - CPAPER
TI  - Deep Learning for Synthesis of Head-Related Transfer Functions
AU  - Bharitkar, Sunil G.
T2  - AES Convention 146
M1  - Paper 10161
PY  - 2019
DA  - 2019/03/06
UR  - https://aes.org/publications/elibrary-page/?id=20294
PB  - Audio Engineering Society
LA  - en
AB  - Ipsilateral and contralateral head-related transfer functions (HRTF) are used for creating the perception of a virtual sound source at a virtual location. Publicly available databases use a subset of a full-grid of angular directions due to time and complexity to acquire and deconvolve responses. In this paper we compare and contrast subspace-based techniques for reconstructing HRTFs at arbitrary directions for a sparse dataset (e.g., IRCAM-Listen HRTF database) using (i) hybrid-based (combined linear and nonlinear) principal component analysis (PCA)+fully-connected neural network (FCNN), and (ii) a fully nonlinear (viz., deep learning based) Autoencoder (AE) approach. The results from the AE-based approach show improvement over the hybrid approach, in both objective and subjective tests, and we validate the AE-based approach on the MIT dataset.
ER  -