Predicting binaural room impulse responses from running signals using a cepstrum-based auditory model and deep learning

Tyler J, Si M, Braasch J. Predicting binaural room impulse responses from running signals using a cepstrum-based auditory model and deep learning. In: AES Convention 155. Audio Engineering Society; 2023. Paper 137. Available from: https://aes.org/publications/elibrary-page/?id=22291

@inproceedings{Tyler2023_22291,
  author = {Tyler, Jeramey and Si, Mei and Braasch, Jonas},
  title = {{Predicting binaural room impulse responses from running signals using a cepstrum-based auditory model and deep learning}},
  booktitle = {AES Convention 155},
  note = {Paper 137},
  year = {2023},
  month = oct,
  publisher = {Audio Engineering Society},
  url = {https://aes.org/publications/elibrary-page/?id=22291}
}

TY  - CPAPER
TI  - Predicting binaural room impulse responses from running signals using a cepstrum-based auditory model and deep learning
AU  - Tyler, Jeramey
AU  - Si, Mei
AU  - Braasch, Jonas
T2  - AES Convention 155
M1  - Paper 137
PY  - 2023
DA  - 2023/10/06
UR  - https://aes.org/publications/elibrary-page/?id=22291
PB  - Audio Engineering Society
LA  - en
AB  - An acoustic model for predicting room features from a running binaural signal is proposed in this study. The spatial locations of the direct sound source and its early reflections are extracted by training a convolutional neural network on a precedence effect model. The precedence effect model uses cepstral analysis, logarithmic filters, cross-correlation, and deconvolution. Using various signals, a synthetic collection of binaural signals was created. The binaural model generates binaural activity maps to from binaural input signals, which are subsequently utilized to train a convolutional neural network. The capacity to forecast the degree of sidedness of a direct sound source and its reflections as well as the reflection delays, has academic applications such as perceptual modeling and room acoustical analysis.
ER  -