Sound Source Identification Method Based On Cross-spectrum Eigenvalue Reconstruction And Elastic Net Beamforming

Beamforming sound source identification technology based on microphone array for measuring signals is applied in many fields such as automobiles and aerospace due to its simple measurement,high computational efficiency and high spatial resolution at high frequency.Beamforming technology is suitable for the measurement of far-field and highfrequency sound sources.Therefore,the test is usually carried out in a cluttered or relatively open space,and the various noises generated in these complex environments will make the reconstructed sound field imaging result have a large number of side lobes,and even the wrong sound field distribution,which makes it impossible to accurately identify and locate the target sound source.Therefore,how to effectively suppress the influence of interference noise on the performance of beamforming sound source recognition is one of the research hotspots in the field of sound source recognition.Aiming at this problem,the paper proposes an elastic net regularized beamforming method(i ENBF)to improve sound source identification.First of all,assuming that the sound wave is a plane wave,according to the traditional delay and summation cross-spectrum beamforming theory,the corresponding algorithm program is written,and its shortcomings are analyzed through numerical simulation,that is,the low-frequency spatial resolution is poor and the high-frequency sidelobe level is high.In order to suppress channel-noise,the self-spectrum method and the self-spectrum reconstruction method are used.The former will cause the reconstructed sound field to appear negative power and produce wrong sound field distribution;the latter uses the semi-positive definiteness of the CSM to make the sound source identification and localization more accurate.However,the interference noise includes not only channelnoise but also background noise.Therefore,the paper chooses the “prewhitening” method and the eigenvalue equally weighted method(EEWM)method to reconstruct the eigenvalues of the CSM,and to ensure the semi-positive definiteness of the matrix.Through numerical simulation analysis,these algorithms can effectively attenuate the sidelobe to suppress the influence of interference noise,and they also can accurately locate the sound source,especially the noise reduction performance of the EEWM.Further,in order to improve the noise reduction effect of the above method,an elastic net regularized beamforming method improved(i ENBF)sound source identification is proposed.This method uses the beamforming output results denoised by the selfspectrum reconstruction method and the eigenvalue improvement method as a priori information,which is constructed into a beamforming regularization matrix to accurately“punish” the source intensity distribution,and combined with elastic net regularization beamforming(ENBF)algorithm.Through numerical simulation analysis of different sound source types,the improved elastic net regularized beamforming method not only has stronger attenuation sidelobe suppression noise ability,higher spatial resolution,and wider analysis frequency range.Finally,the practicability and feasibility of the sound source identification method of i ENBF is verified through the experiment of different sound sources,which provides a reference for practical engineering applications.