Research On Improved Beamforming Algorithm Based On Phase And Frequency Conversion

In recent years,it is the rapid development of Chinese high-speed railway that makes the problem of pollution in this field is becoming more and more prominent.The noise sources are studied firstly to make control and governance over environmental noise,and the necessary mean is noise source identification technology.At present,conventional sound source identification technology is aimed at medium and high frequency sound sources,but few studies have been done on low frequency noise,which is an important component of high-speed railway noise.The technical bottleneck lies in the defect of the algorithm and the limitation of array size in practical testing.Based on this situation,this paper starts from the traditional beamforming technology to improve its algorithmic performance.Based on the principle that microphones receive different phase difference from low-frequency and high-frequency sources,the phase difference which exist from low-frequency sound source to microphones are compensated artificially to make low-frequency be identified as a high-frequency sound source.Specific research is as follows:Firstly,the basic theory of sound source identification is deduced.The models of narrowband and broadband signals received by array are introduced.The theoretical deductions of traditional beamforming,cross-spectrum-based beamforming algorithm and DAMAS inverse operation are given.The relationship between the two beamforming algorithms with the frequency of sound source is studied by simulation.Secondly,two deconvolution algorithms with good performance are screened out,and the principles of the two algorithms are deduced,the comparative simulation of the two algorithms is carried out.The results show that the recognition result of FISTA with good robustness is better than that of FFT-NNLS algorithm.Both algorithms can identify coherent and incoherent sources,while there is a deviation over amplitude when identifying coherent sources.The amplitude error of the two algorithms is larger than that of the narrow band when they are used to identify broadband sources.Then,the shortcomings of deconvolution algorithm and the necessity of phase-frequency transformation are presented.Two schemes of phase-frequency transformation are proposed.The second method can identify single and double incoherent sources.In order to solve the phase-frequency conversion problem of coherent multi-sound sources,the signal unwrapping of microphone using CVX toolbox is proposed,subsequently,the phase-frequency conversion is carried out with the first method,or using vector rotation method to finish the operation of phase-frequency conversion.However,all of these methods have their shortcomings in application.Over this basis,the causes of phase-frequency conversion are further analyzed,and the method of enlarging phase difference of microphone is proposed by array virtual technology,and the simulation test is carried out later.The method of correcting the additional phase deviation is proposed from the two aspects of experiment and algorithm,which makes the phase-frequency transformation of low-frequency sound source more accurate.The entire operation is considered to be significant in engineering application.Finally,the static sound source identification experiment outdoors is carried out,and all the algorithms are tested in practice.The results show that the algorithm and program in this paper are correct.By using Doppler time-domain correction method,the recognition of stationary sound source is extended to the field of moving sound source.In addition,the post-processing software of laboratory is designed based on VB and MATLAB mixed programming which could analyze the data received by array.