Generative Adversarial Networks for Audio Equalization: an evaluation study

Pepe G, Gabrielli L, Squartini S, Cattani L, Tripodi C. Generative Adversarial Networks for Audio Equalization: an evaluation study. In: AES Convention 148. Audio Engineering Society; 2020. Paper 10367. Available from: https://aes.org/publications/elibrary-page/?id=20784

@inproceedings{Pepe2020_20784,
  author = {Pepe, Giovanni and Gabrielli, Leonardo and Squartini, Stefano and Cattani, Luca and Tripodi, Carlo},
  title = {{Generative Adversarial Networks for Audio Equalization: an evaluation study}},
  booktitle = {AES Convention 148},
  note = {Paper 10367},
  year = {2020},
  month = may,
  publisher = {Audio Engineering Society},
  url = {https://aes.org/publications/elibrary-page/?id=20784}
}

TY  - CPAPER
TI  - Generative Adversarial Networks for Audio Equalization: an evaluation study
AU  - Pepe, Giovanni
AU  - Gabrielli, Leonardo
AU  - Squartini, Stefano
AU  - Cattani, Luca
AU  - Tripodi, Carlo
T2  - AES Convention 148
M1  - Paper 10367
PY  - 2020
DA  - 2020/05/06
UR  - https://aes.org/publications/elibrary-page/?id=20784
PB  - Audio Engineering Society
LA  - en
AB  - In this paper we propose a neural network-based approach for audio equalization inside a car cabin. We consider the Generative Adversarial approach to generate FIR filters for binaural equalization at the driver listening position of the sound produced by multiple loudspeakers. The neural network is optimized to generate equalizing filters able to achieve a flat frequency response at one control position in a time-invariant scenario. Results are analyzed in the frequency domain, comparing the achieved frequency response with the desired one. Compared to previous works, the proposed approach provides better results with a very low error compared to the target response.
ER  -