Surface Enhanced Raman Scattering Spectrum Recognition for Trace Detection of Common Drugs in Urine / 分析化学

ABSTRACT

Assembling an adapted smoothing method and a classifier of wavelet transform combined support vector machine ( SVM) , a Raman spectrum recognition approach was built for low signal noise ratio situation. Firstly, spectra data were denoised by the adapted smoothing method. The smoothing window was adapted to the signal noise ratio, which would effectively remove noise with the intensity of the signal well remained. Secondly, the wavelet transform was used for dimension reduction of the data. The decomposition level of wavelet transform was optimized according to the best classification result of the training set. Lastly, SVM was used for classification. Cross Validation ( CV ) was applied to obtain the optimized parameters of SVM. Conditions for the effective parameters were searched considering the relation between the cross_validation result and the classification accuracy. Combined with the surface enhanced Raman scattering ( SERS ) technology , the developed spectrum recognition approach was used for qualitative analysis of methamphetamine ( MAMP ) and 3, 4_methylenedioxymethamphetamine ( MDMA ) in peopleˊs urine, where the detecting accuracy is above 95. 0%. The uniform Au nanorods (NRs) SERS substrate synthetized by the Hefei Institute of Intelligent Machines of Chinese Academy of Sciences was used for the experiment. Raman spectra were acquired using an Inspector Raman ( DeltaNu) spectrometer, with the excitation wavelength of 785 nm and the integrate time of 5 seconds.