Ghost Image Suppression Based on Particle Swarm Optimization-MVDR in Sound Field Reconstruction

Abstract

In sound field reconstruction, spurious sources called ghost images always appear around the position of the real sound source in the sound pressure distribution map because of the grating and side lobes, thus resulting in an incorrect identification of the sound source. To solve this problem, a method for suppressing ghost images is proposed in this paper; such method is based on particle swarm optimization (PSO) and minimum variance distortionless response (MVDR) beamforming. In this method, the elements distribution of a microphone array is first optimized by the PSO algorithm to acquire the optimal design of an unequal spacing microphone array. With this array, the grating lobe is suppressed, and the increscent value of the inherent side lobe value is reduced. Second, MVDR algorithm is used to weaken the effect of the side lobes and to obtain a sound pressure distribution map in which the ghost images are suppressed. The advantage of this method is the combination of the unequal spacing array, which suppresses the grating lobe, and the MVDR algorithm, which has excellent performance in spatial filtering. Through this method, a microphone array with a few number of elements can achieve ghost image suppression. Experiments on sound field reconstruction in an anechoic chamber for a single-tone sound source are conducted to validate the proposed method. Moreover, some extra sound field reconstructions for a single-tone sound source and double sound sources with broadband in a normal room with different parameters such as the array shape and distance from the sources to the array are conducted to discuss their influences on the effectiveness of the proposed method.