Microphone Cross-talk Cancellation in Ensemble Recordings with Maximum Likelihood Estimation

Das O, Smith IJO, Abel JS. Microphone Cross-talk Cancellation in Ensemble Recordings with Maximum Likelihood Estimation. In: AES Conference: 2020 AES International Conference on Audio for Virtual and Augmented Reality (August 2020). Audio Engineering Society; 2020. Paper 10471. Available from: https://aes.org/publications/elibrary-page/?id=21148

@inproceedings{Das2020_21148,
  author = {Das, Orchisama and Smith, III, Julius O. and Abel, Jonathan S.},
  title = {{Microphone Cross-talk Cancellation in Ensemble Recordings with Maximum Likelihood Estimation}},
  booktitle = {AES Conference: 2020 AES International Conference on Audio for Virtual and Augmented Reality (August 2020)},
  note = {Paper 10471},
  year = {2020},
  month = aug,
  publisher = {Audio Engineering Society},
  url = {https://aes.org/publications/elibrary-page/?id=21148}
}

TY  - CPAPER
TI  - Microphone Cross-talk Cancellation in Ensemble Recordings with Maximum Likelihood Estimation
AU  - Das, Orchisama
AU  - Smith, III, Julius O.
AU  - Abel, Jonathan S.
T2  - AES Conference: 2020 AES International Conference on Audio for Virtual and Augmented Reality (August 2020)
M1  - Paper 10471
PY  - 2020
DA  - 2020/08/06
UR  - https://aes.org/publications/elibrary-page/?id=21148
PB  - Audio Engineering Society
LA  - en
AB  - While recording an ensemble of musicians, it is often desired to isolate the instruments to avoid interference from other sources. Close-miking and acoustic isolation booths are some techniques for mitigating microphone cross-talk (or “bleed"). In this paper, we propose an algorithm for canceling microphone bleed in ensemble recordings in the post-processing stage. We propose a calibration stage to estimate the relative transfer function from each instrument to each mic. Then, we set up an optimization problem to simultaneously estimate the transfer functions and individual sources given the microphone signals and a noisy estimate of the transfer function obtained from the calibration stage. We show that minimizing this cost function gives us the maximum likelihood estimate when we assume the distributions to be normal. Finally, we test our proposed method to cancel microphone bleed in a synthesized environment, and compare our results to an existing multichannel Wiener filter method.
ER  -