Research On Moving Noise Source Localization Methods Based On Plane Microphone Array

The Noise source identification is the premise condition of noisesuppression and reduction. Due to the type, frequency and other characteristics ofhigh-speed motion noise will vary with the train working status and rates, it isdifficult to measure and model the external noise. Classic delay-sum beamformingalgorithm can’t overcome Doppler Effect in motion noise source. Therefore, it is ofgreat significance to study the theoretical, design array and simulate the subject ofidentification the sound source of motion noise.Combined with project of research on identification of the sound source ofmotion noise, based on the previous domestic and international research experience,simulation research and theoretical approaches to identify motion noise are carriedout. This paper mainly includes the following several parts.(1) The defects of motion sound source identification using classic delay-sumbeamforming algorithm are put forward, on the basis of this method adopted instatic sound source identification. The movement of sound radiation laws areelaborated, sound field radiation mathematical description is given, the regularityand characteristics are deduced.(2) Moving sound source point and measurement point geometry model is setup, on the foundation of elementary radiation principles and beamforming methodof moving sound sources, the basic analysis theory of array signal for high-speedtrain is described. Focusing on the Doppler effect still remains in wave numberdomain and the simplified model experimental error, the de-Dopplerizationprocedure is studied through differential and integral calculus, then, the motionsource sound field reconstruction formula is deduced on the basis of non-simplifiedmodel. In order to improve the property of time domain beamforming method theautocorrelation of microphone signal is removed.(3) The geometry structure of microphone array shape is important for theirperformance, and plays the key factors impacting for external noise sourceidentification. The simulation of the array directivity, angle resolution and otherperformance of cross-array and hexagonal-array (37elements), rectangular-array(36elements) are performed and analyzed in qualitatively and quantitatively. (4) Based on azimuth spectrum estimation theory, the conventionalbeamforming, MVDR beamforming, MUSIC beam shaper azimuth spectrum areshown when two sound source azimuth interval is5°and10°. Throughcomparing the relationship between resolution and interval, SINR and SNR,stability, array gain and other performance of those beamforming are studied, whichprovides a basis algorithm selection for subsequent sound field reconstruction.(5) In view of the relevant conclusions of microphone array selection inpractical noise source identification and the comparative study of beamformingalgorithm，utilize MATLAB tool, combine with conventional beamforming, MVDRbeamforming algorithm to reconstruct the moving noise source field forsingle-frequency sound source, multi-frequency sound source, dipole anddistribution source based on the non-simplified model and simplified model.