Simulation investigation of spatial interpolations in virtual rotating array beamforming with different array configurations for rotating sound source localization

Abstract

Virtual rotating array (VRA) beamforming is a frequency domain rotating sound source localization algorithm, whose basic concept is to make the microphones stationary relative to the sound source. Previously, VRA beamforming was restricted to the array configuration of ring formation. Recently, scholars have successfully applied several spatial interpolation methods to an array with arbitrary microphone configuration for VRA beamforming. However, the performance of these interpolation methods in different array configurations has not been verified. In this paper, spatial interpolation methods in VRA beamforming are applied to several array configurations by simulation, to verify their Doppler shift compensation capability, spectrum reconstruction capability, sound source position precision and sound source strength accuracy. Meanwhile, comparing the results with those of a ring array. Results show that regardless of the interpolation method and array configuration applied, the Doppler effect can be compensated well. The performance of interpolation methods is affected by the array configuration, yet the inverse distance weighting class performs well in terms of position precision across all arbitrary array configuration. Although each arbitrary array configuration can better locate all sound source locations, it cannot accurately identify the sound intensity differences of each source. While the ring array can identify the position and sound intensity of each source more accurately.