Influence of Radius of Microphone Array on Virtual Rotating Array Beamforming for Rotating Noise Source Localization

The Virtual Rotating Array (VRA) beamforming algorithm is an efficient algorithm for localizing rotating noise sources, especially with ring arrays in the frequency domain. VRA has now been extended to work with arbitrary microphone arrays. Determining the proper array radius and its impact on VRA beamforming is still a major challenge in installing arbitrary microphone arrays for the better achievement of beamforming results. The purpose of this paper is to investigate the influence of arbitrary microphone arrays with different radii on virtual rotational beamforming. The arbitrary microphone configurations employed are the Underbrink, H0V5, and H0V3.13 arrays. The results show that at frequencies within 2.5 ≥ f ≥ 5 kHz, there is usually a source intensity error when the array radius is half the size of the rotating source radius making other sources hard to identify. This may be due to array element spacing error or Rayleigh’s criterion. Furthermore, localization results are significantly enhanced when the radius ratio (array radius/rotating source radius) is within 2–4 at frequencies(f) within 5 ≥ f ≥ 8 kHz. Finally, employing deconvolution algorithms like CLEAN-SC for VRA beamforming greatly improves spatial resolution and source location precision.