Improved algorithm of three-dimensional beamforming based on spatial cross-array

Abstract

Three-dimensional beamforming based on microphone array signal processing is the expansion of traditional 2D beamforming. However, its identification accuracy is often badly reduced by the effect of grating lobes and side lobes. To overcome this problem, a method called the hybrid method beamforming (HMB) combining functional generalized inverse beamforming with multiplicative filter of spatial cross-array is proposed. In this method, the statistically optimal processing and iterated generalized inverse beamforming with regularized matrix function are utilized to obtain initial result. Then the high order function is applied to filter the output. Subsequently, a novel non-uniform spatial cross-array optimized by genetic algorithm is used to obtain sound pressure distribution. The array consists of three orthogonal sub-arrays. Mutual cancellation is realized by computing respectively with data of sub-arrays and multiplying together. With fewer microphones, the result of the improved method can be obtained with a higher spatial resolution. The proposed method is verified by the simulation and the source localization test in a room. Compared with the conventional frequency domain beamforming (FDBF) algorithm and statistically optimal array processing (SOAP) beamforming, the performance of the proposed method is significantly improved in terms of resolution of the acoustic source.