A New Performance Criterion for Microphone Array Geometries and Filterand- Sum Beamforming

Abstract:
Our aim is to quantify microphone array performance with respect to its geometry, when the array is used for wideband beamforming. To this end we present a new array– geometry performance criterion which takes the noise reduction capability and the beamformer robustness to spatially uncorrelated white noise into account. The latter aspect is especially important in practical implementations as it is often the cause of low-frequency noise boost in the beamformed signal. We next simulate various linear array configurations and use the proposed measure to indicate optimal configurations for a wideband minimum variance distortionless response (MVDR) beamformer. The obtained optimal arrays are analyzed and compared to the widely used uniform linear array configurations. Results show that the optimal arrays in most cases outperform the uniformlinear arrays in terms of robustness to spatially uncorrelated noise as well as in noise suppression capability.