A frequency-domain beamforming for rotating sound source identification.

Abstract

Identifications of rotating sound sources are of interest in many industrial applications. Nowadays, almost all frequency-domain rotating beamformings are based on the concept of virtual rotating array (VRA), and consequently, have one constraint that microphones must constitute a ring around the rotation axis of a sound source. This constraint greatly restricts the wide application of these beamformings in industrial experiments. To overcome this constraint, this paper proposed a frequency-domain rotating beamforming (mode composition beamforming, MCB). Without using the concept of VRA, MCB directly uses the concept of delay-and-sum with the mode composition of rotating sound. MCB has the same ability to rotate sound source identification as the famous time-domain rotating source identifier (ROSI) proposed by Sijtsma et al. (2001) and generally has a much higher computational speed than ROSI, especially when a large number of samplers are used. Special arrays, such as rotational symmetry array, with respect to the rotation axis with a high order of rotational symmetry, are recommended to compensate for the ghost contributions in MCB induced by the Doppler effect.