Beamforming Methods Based On Spectral Matrix Decomposition And Reconstruction For Acoustic Source Identification

The sound source identification method of beamforming based on microphone array has been widely applied in many acoustic source identification fields such as automobile and aerospace because of its advantages of simple measuring, fast calculation speed and good resolution in high frequency. In order to obtain more complete and accurate results of sound source identification, the emphasis of the research is a series of clearness beamforming algorithms, especially deconvolution algorithm. However, it is presently hard to find research about the clearness method based on spectral matrix decomposition and reconstruction. This paper mainly researching and analyzing two kinds of spectral decomposition and reconstruction beamforming algorithms entirely, and that using it to identify the weakness of car dash panel insulation successfully.First of all, based on the spherical wave assumption, the theories of cross-spectral delay-and-sum beamforming and cross-spectral imaging function were proposed in detail and the performance evaluation indexes were expounded briefly, and then identification images of single source and multiple sources were simulated. The results show that both algorithms can identify different types of source effectively and the higher frequency, the better resolution. However, the identification image is polluted by high sidelobes and the identification accuracy is influenced by wide main lobe. when there is weak source below the maximum side lobe, then the conventional beamforming method is unable to identify the weak source for its limits.To overcome the shortcomings in conventional beamforming method, the singular value decomposition beamforming was proposed and the identification images of strong and weak sources were simulated. The results show that this method not only identify the strong source but also it breaks through the limit of side lobe to identify the weak source exactly for high frequency or non-fusion incoherent sources. Whereas, this method can't be used to identify coherent sources and doesn't have the ability to improve resolution or reduce side lobe. Based on the incoherent sources example experiment were conducted to validate the correctness of simulation and the effectiveness of the singular value decomposition beamforming in the practical application.Furthermore, on basis of the singular value decomposition, the functional beamforming method was given to obtain clearer images. The simulation results of single source and multiple sources show that his method can reduce side lobe effectively and improve resolution slightly, ideally, the identification images looks like clearer by lager exponent. However, the output error of main lobe increase with the increase of exponent under non ideal condition. According to a great of simulation results, the error curves of outputs were drawn and 16 was the suggested exponent value. On this basis, the four incoherent speaker sources example experiment were conducted to validate the correctness of simulation and the ability of weak source identification.In the end, the experiment of identifying weak sound insulation part of car dash panel was done. The air-conditioning air inlet is the weak sound insulation part of car dash panel based on functional beamforming method. The sound insulation performance of car dash panel was improved by pastes sealing material between the inner and outer circulation valve and its port.