Parametric Spatial Filter Utilizing Dual Beamformer and SNR-Based Smoothing

Delikaris-Manias S, Pulkki V. Parametric Spatial Filter Utilizing Dual Beamformer and SNR-Based Smoothing. In: AES Conference: 55th International Conference: Spatial Audio. Audio Engineering Society; 2014. Paper 1-2. Available from: https://aes.org/publications/elibrary-page/?id=17369

@inproceedings{DelikarisManias2014_17369,
  author = {Delikaris-Manias, Symeon and Pulkki, Ville},
  title = {{Parametric Spatial Filter Utilizing Dual Beamformer and SNR-Based Smoothing}},
  booktitle = {AES Conference: 55th International Conference: Spatial Audio},
  note = {Paper 1-2},
  year = {2014},
  month = aug,
  publisher = {Audio Engineering Society},
  url = {https://aes.org/publications/elibrary-page/?id=17369}
}

TY  - CPAPER
TI  - Parametric Spatial Filter Utilizing Dual Beamformer and SNR-Based Smoothing
AU  - Delikaris-Manias, Symeon
AU  - Pulkki, Ville
T2  - AES Conference: 55th International Conference: Spatial Audio
M1  - Paper 1-2
PY  - 2014
DA  - 2014/08/06
UR  - https://aes.org/publications/elibrary-page/?id=17369
PB  - Audio Engineering Society
LA  - en
AB  - This paper proposes a dual-beamformer implementation of the cross pattern coherence algorithm (CroPaC). CroPaC is a parametric spatial filtering technique that utilizes coherence-based measures between signals from microphones of different orders as a criterion for focusing in specific directions. A spatial filter is calculated from the multichannel input of a microphone array and it assigns attenuation values to the output of a minimum variance distortionless response beamformer. An instantaneous signal-to-noise ratio based adjustment of the smoothing parameters is proposed for the spatial filter to mitigate modulation artifacts such as musical noise. A prototype microphone array is built and tested in a multi-talker scenario with several background noise levels. Results show an improvement in terms of noise reduction and attenuation of distortion when compared to the original version of the algorithm.
ER  -