Learning Filters in Feedback Delay Networks From Noisy Room Impulse Responses

Dal Santo G, Prawda K, Schlecht SJ, Välimäki V. Learning Filters in Feedback Delay Networks From Noisy Room Impulse Responses. J Audio Eng Soc. 2026;74(10):731-749. doi:10.17743/jaes.2026.0294

@article{DalSanto2026_23444,
  author = {Dal Santo, Gloria and Prawda, Karolina and Schlecht, Sebastian J. and Välimäki, Vesa},
  title = {{Learning Filters in Feedback Delay Networks From Noisy Room Impulse Responses}},
  journal = {Journal of the Audio Engineering Society},
  volume = {74},
  number = {10},
  pages = {731--749},
  year = {2026},
  month = oct,
  publisher = {Audio Engineering Society},
  doi = {10.17743/jaes.2026.0294},
  url = {https://doi.org/10.17743/jaes.2026.0294}
}

TY  - JOUR
TI  - Learning Filters in Feedback Delay Networks From Noisy Room Impulse Responses
AU  - Dal Santo, Gloria
AU  - Prawda, Karolina
AU  - Schlecht, Sebastian J.
AU  - Välimäki, Vesa
T2  - Journal of the Audio Engineering Society
J2  - J. Audio Eng. Soc.
VL  - 74
IS  - 10
SP  - 731
EP  - 749
PY  - 2026
DA  - 2026/10/09
DO  - 10.17743/jaes.2026.0294
UR  - https://doi.org/10.17743/jaes.2026.0294
PB  - Audio Engineering Society
LA  - en
AB  - Recursion is a fundamental concept in the design of filters and audio systems. In particular, artificial reverberation systems that use delay networks depend on recursive paths to control both echo density and the decay rate of modal components. The differentiable digital signal processing framework has shown promise in automatically tuning recursive and nonrecursive elements using gradient-based optimization with perceptually or physically motivated loss functions, such as energy decay or spectrogram differences. These representations are highly sensitive to model mismatches, which can lead to spurious loss minima. In particular, discrepancies in background noise can result in inaccurate attenuation estimates. This paper addresses the problem of tuning recursive attenuation filters of a feedback delay network when targets are noisy. The loss profile associated with different optimization objectives is analyzed, and a method that explicitly models noise is proposed, which improves the accuracy of the estimated attenuation filters under low signal-to-noise conditions. The effectiveness of the proposed approach is demonstrated through statistical analysis on both synthetic and real target data. Furthermore, the sensitivity of attenuation filter parameters tuning to perturbations in frequency-independent parameters is identified. These findings provide practical guidelines for more robust and reproducible gradient-based optimization of feedback delay networks.
ER  -