Spatial Filtering
Shmulik Markovich‐Golan, Walter Kellermann, Sharon Gannot · 2018
This chapter is dedicated to spatial filtering, that is the application of a linear, often time-varying filter called beamformer to a multichannel signal acquired by a microphone array. We introduce some fundamentals of microphone array processing, including the signal model and the characterization of spatial selectivity by beampatterns, directivity and sensitivity. We then discuss array topologies followed by data-independent beamforming and data-dependent beamforming. For a general treatment of the latter, we review the use of relative transfer functions (RTFs) and various design criteria. Data-dependent beamforming is then related to binary masking and also to the spatial filtering properties of blind source separation (BSS) algorithms, and the use of BSS algorithms as a component within data-dependent beamformers is delineated. We present a predominant structure for realizing data-dependent beamformers, the generalized sidelobe canceler (GSC). Finally, we address the special role of single-channel postfiltering as a complement to multichannel spatial filtering.