Cross spectral density based spatial filter employing maximum directivity beam patterns

Abstract

This paper proposes a spatial filter based on the cross-spectral density between directional microphones. The operation principle of the proposed spatial filter is predicated on the principles of the cross pattern coherence, which is a parametric spatial filtering technique that uses coherence-based measures as a criterion for focusing on specific directions. In this work, the choice of the directional microphones is performed according to the microphone array specifications. A spatial filter is then calculated and assigns attenuation values to the output of a beamformer. In the experimental validation, where a room simulation is used to generate a virtual recording scenario, the spatial filter provides an improvement in terms of interference suppression over conventional beamforming.