Research On Generalized Inverse Sound Source Identification Method Based On The Microphone Array

The accurate location and identification of the noise source is the precondition of noise control.Sound source identification method based on microphone array can detect the sound source effectively and realize the visualization of sound field.So it has become the focal and hot themes of the NVH field.Since the actual noise source in the engineering measurement are broad-band sources in general.To identify the noise source accurately,the source identification method must be able to reconstruct the sound pressure,particle velocity at any point of the source surface in wide frequency bands.For the identification of multi-sources in a complicated environment,the source distribution is susceptible to sidelobe ghosting caused by the interference noise.Only the source identification methods have a wide dynamic display range,the location of the noise source can be accurately identified.Therefore,how to effectively compensate for the lack of the source identification methods at frequency bandwidth,spatial resolution and dynamic range,has become the hotspot of the sound source identification.To improve the spatial resolution and the dynamic range of acoustic holography and beamforming,two improved generalized inverse source identification method are introduced.After introducing the above methods,they can be used to locate the sound source in the structural plate and interior noise source of the vehicle.Based on the research of general source identification inverse problems,combining the generalized inverse source identification method and the regularization matrix and scale parameter,an improved generalized inverse source identification method is proposed to reduce the effect of reconstruction and measurement error.Using threshold filtering to suppress the sidelobe ghost during the inverse process,it can ensure the recognition and reconstruction precision.Using a simple source model for the actual establishment of sound source identification,the performances of the improved generalized inverse source identification method are explored under the condition of the different distance measurement and signal to noise ratio(SNR)by the numerical simulation.The results show that the improved generalized inverse source identification method has high resolution and good robustness in a certain measuring distance.At the same time,for monopole and coherent sources,the numerical simulation and experiments show that the improved generalized inverse source identification method can accurately identify the coherent sources although there are strong interference sources.In the experiment,combined the improved generalized inverse source identification method with sound quality estimation,it can realize the reconstruction of the sound quality parameters for the broadband source on the source plane.It also shows that the improved generalized inverse source identification method can not only realise the source identification,but also can estimate sound quality parameters.Base on the cross spectral beamforming and generalized inverse source identification method,introducing a high order function of the pressure cross matrix and normalized steering vectors,the high order function generalized inverse source identification method is proposed to achieve non-destructive source intensity for the multi-sources identification.The high order output can reduce the width of the main lobe and restrain sidelobes.So it improve the dynamic range of the sound source identification.Using point source reconstruction theory,the independence condition of eliminating intensity estimation error is introduced and the amplitude correction is proposed.The convergence analysis also proves that multi-sound source can be identified.Therefore,the relationship between the reconstruction accuracy and order is analyzed by means of numerical simulation under different frequencies and SNR.It is found that its spatial resolution can be improved by increasing order.Through the above analysis,the order v range is recommended and order v can be limited to 1-20.At the same time,through numerical simulation and experiment,it is found the high order function generalized inverse source identification method has high resolution and wide dynamic range relative to other source identification methods.Under idle condition,the automotive engine test also verifies that the high order function generalized inverse source identification method can accurately identifiy the main source of motivation and has better dynamic performance.Considering the existence of various reflections for the vehicle interior source identification,combining two proposed source identification method and double microphone array,a new mixed sound recognition method is proposed to separate the back sources.Based on double microphone array,the vibration velocity can be reconstructed by the improved generalized inverse source identification method.Also the sound pressure can be reconstructed by the high order function generalized inverse source identification method.Based on the numerical simulation analysis,it is found that the optimized method can improve the performance of source separation by increasing order.At the same time,it also found the front main source can accurately separated by the optimization method when the amplitude of front and back sources is inconsistent.Also its dynamic range is wide.The test for the actual loudspeaker source also show that the proposed method is effective.By introducing high order cross matrix and Green function with evanescent wave into the source identification model,a new method is proposed to conduct identification research for the structural panels by combining two improved a generalized inverse source identification method and inverse Pacth method.This method fully inherited the advantage of high order generalized inverse source identification method,it also can has high precision reconstruction compared to the iPTF method.By introducting two kinds of Green function,it discusses the relationship between the reconstruction accuracy and the condition number of the Green function.Taking into account the Green function based on evanescent wave has high computational efficiency and also can accurately reflect the evanescent wave transmission.By the numerical simulation for the piston source and the vibration of structural panels,it found the improved generalized inverse source identification method has high recognition performance for the sparse and non-sparse source.Also it can improve the source localization ability by adjusting the order.The test results for identifying the real plate source further verify the feasibility of the proposed method.