Signal-Dependent Mixing for Direction-Preserving Multichannel Noise Reduction

Adrian Herzog and Emanuël A. P. Habets

Submitted to EUSIPCO Conference 2021.


Reducing undesired sounds in a multichannel recording while preserving the spatial characteristics of the acoustic scene is a challenging task. In recent works by the present authors, different Ambisonics-to-Ambisonics noise reduction methods were discussed. To mitigate spatial distortions of the noise, a direction-preserving approach and a partial-noise-reduction approach were investigated. In this work, a signal-dependent partial-noise-reduction approach is proposed, which is able to mitigate spatial distortions of the noise further without significantly deteriorating other performance measures and can be applied to any multichannel signal format. The proposed method is evaluated using spherical-microphone-array recordings from the ACE corpus.


Binauralized third-order Ambisonic signals containing speech and noise.
The mixing factor for the partial noise reduction methods and the lower bound for the direction method were set to -20 dB. For theproposed signal-dependent partial noise reduction method, the noise distortion trade-off parameter was set to 8.

Files were generated using the SPARTA AmbiBIN VST plugin [1] with default HRIRs.
Playback with headphones is recommended.

For the reverberant signals, impulse responses and noise signals from the ACE corpus [2] were used.

Anechoic speech and diffuse white noise (not evaluated in paper).
Reverberant speech and babble noise.
Reverberant speech and fan noise.
Reverberant speech and ambient noise.


[1] L. McCormack and A. Politis - SPARTA and COMPASS: Real-time implementations of linear and parametric spatial audio reproduction and processing methods, AES Conf. Immersive and Interactive Audio, York, UK, March 2019. [2] J. Eaton, N. D. Gabuich, A. H. Moore and P. Naylor - The ACE challenge - corpus descriotion and performance evaluation, in Proc. IEEE Workshop on Appl. of Signal Process. to Audio and Acoust. (WASPAA), New Paltz, NY, USA, Oct. 2015.