The Research Of Noise Source Identification Based On Generalized Inverse Beamforming Via TSVD Regularization

Continuous increase of train speed makes noise pollution of high-speed rail increasingly serious.This hampered railway transportation speed up,so the control of noise of high-speed rail becomes more and more important.It is a key factor to find the reason for the noise accurately to control the noise of high-speed rail,so sound sources should be identified firstly.To identify sound sources for complex extended noise sources,this paper proposes a method based on array miniaturization and singular truncation regularization generalized inverse beamforming(TSVD-GIB)algorithm,then designs the sound source identification system based on the method,and set up a 4×4 rectangular microphone array.Finally we carry out some simulation and experiments to test the method's effectiveness.In this paper,main work and innovations are as follows:(1)Explaine the working principle systematically of the linear array and plane array beamforming,then analyze the influence of noise signal frequency,array spacing,number of sensors and array size to the beam scanning resolution by numerical simulation.(2)Analyze the principle and implementation process of GIB algorithm by generalized inverse operation.Establishe extended sound sources model,and compare the resolution of conventional beamforming(CBF)algorithm and GIB algorithm of point source,extended sources(two point sources,three random distribution of point sources)by numerical simulation.The results show that GIB algorithm can effectively identify point sources and extended sources.However GIB algorithm get a low scanning resolution and CBF algorithm is completely failed when the distance between sound sources is 14 cm.(3)Propose a method based on singular truncation regularization of generalized inverse beamforming.Compare scanning resolutions of CBF algorithm,GIB algorithm and TSVD-GIB algorithm on the identification of point sources and extended sources(5 point sources).The results show that TSVD-GIB algorithm can effectively identify 4 point sources which distance among each other is 14 cm,and its errors are less than 2 cm,acoustic center position is the smallest compared with CBF algorithm and GIB algorithm.TSVD-GIB effectively eliminate sidelobe interference,improves the accuracy of sound sources identification.(4)Set up rectangular microphone array in the semi-dead room,single speaker and two speakers which distance is 20 cm are used respectively to identify the sound sources.The results show that,for single speaker,all of CBF algorithm,GIB algorithm and TSVD-GIB algorithm can identify sound source effectively,and their identification errors are all less than 5cm.Acoustic center of TSVD-GIB algorithm is the smallest,and its accuracy is the best.Only TSVD-GIB algorithm can identify two speakers with same frequency effectively.Its deviation on x axis and y axis are both less than 4cm.However,CBF algorithm is completely failed and GIB algorithm gets a bigger error.