Application Of Variational Mode Decomposition Algorithm In Adaptive Denoising Of Raman Spectroscop

With the continuous development of Raman spectroscopy detection technology,Raman spectroscopy,as a molecular scattering spectrum with fast analysis speed,non-destructive,non-contact,small detection limit and high selectivity,has been widely used in materials,chemical industry,medical treatment,agriculture and other fields,making real-time,on-site,dynamic and rapid detection is becoming a reality.However,in the actual acquisition of Raman spectra,the signal may be infected by noise in each process from the laser emitter to the receiver.Especially,for the special requirements of special sample on-site detection,such as short-time exposure,low power excitation laser and other factors,the characteristic information carried by Raman spectra is even submerged in a lot of noise,and the weak Raman spectral signal usually becomes difficult to identify.The problem of using traditional denoising algorithms to obtain weak Raman spectra from strong noise background is that it is difficult to find the best parameter configuration,and it needs to spend more time and energy to constantly try various parameter combinations.Thus,it is necessary to deeply study the adaptive denoising method of Raman spectroscopy with strong noise interference.Therefore,in this study,an adaptive Raman spectroscopy denoising method based on Variational Mode Decomposition(VMD)is proposed.Based on the VMD algorithm,the swarm intelligence algorithm is used to optimize the parameters of the VMD algorithm,the finite number of decomposed Intrinsic Mode Function(IMF)are dynamically divided,and the spectral effective information in the high frequency components is extracted,so as to realize more refined adaptive denoising of Raman spectra.The main research contents are as follows:(1)A variational mode decomposition algorithm based on Multi-Island Genetic Algorithm is constructed.Firstly,the optimization ability of Particle Swarm Optimization algorithm,Genetic Algorithm and Multi-Island Genetic Algorithm for VMD parameters is studied,and the optimization algorithm is determined.Then,a fitness function with the minimum envelope entropy value is constructed,which allows the combination of VMD parameters to be searched flexibly according to the actual situation of the signal.Finally,the best parameter combination[ K,α] is assigned to the number of decomposition levels K and the penalty factor α.It avoids excessive decomposition caused by artificial setting of VMD parameters and improves the accuracy of VMD decomposition signals.(2)The decomposed finite modal components are effectively distinguished,furthermore,the Raman spectra with low SNR can be finely denoised.The Raman spectra are decomposed into a finite number of mode components by VMD algorithm,and the correlation coefficient between each modal component and the original signal is calculated.The correlation coefficient threshold was constructed to scientifically and dynamically divide the modal components,and the modal components were sifted into spectral signal components and noise signal components.The wavelet new threshold denoising method is constructed to carry out hierarchical de-noising of modal components,and the wavelet new threshold denoising method is applied to the noise signal components to extract the high frequency feature information of Raman spectrum.Raman spectra are reconstructed by processed noise signal components and spectral signal components,and then Raman spectra with low SNR can be finely denoisde.(3)Raman spectrum denoising system is developed.The traditional denoising method and the adaptive denoising method based on signal decomposition are integrated into the system.The system realizes the denoising processing of Raman spectrum by various methods.This study is based on the Windows platform,with using MATLAB language,App Designer to achieve the development of Raman spectrum denoising system.